# On Intelligence, Swiftness, and the Corruption of Judgment It is a melancholy observation, which the advancement of mechanical arts has rendered impossible to avoid, that we have achieved the power to commit errors with a velocity that leaves no room for the exercise of that faculty which alone distinguishes us from the machines themselves—namely, the capacity to hesitate, to examine, and to doubt. The programmer who receives output before he has finished articulating his requirement stands in a position which nature never intended for rational creatures: he is asked to be answerable for decisions made at a speed that annihilates responsibility itself. Let us be precise in our terms, for precision here is not mere pedantry but a moral necessity. Intelligence, as commonly understood, signifies the capacity to apprehend relations among things, to perceive what is true and what is useful. But wisdom—that harder thing—is something else entirely. Wisdom is the knowledge not merely of what may be done, but of what ought to be done; it is the marriage of understanding with restraint, of capability with judgment. A man may be intelligent and yet lack wisdom; he may discern the solution to a problem with perfect clarity whilst remaining blind to whether the speed of that solution has corrupted the very grounds upon which answers may be trusted. The machine, I grant, is intelligent in its fashion. It perceives patterns in code as swiftly as the human mind perceives a familiar face. But it is wisdom that the machine cannot possess, and wisdom that those of us who deploy it are rapidly abandoning—not from malice, but from a kind of intellectual intoxication, a fascination with velocity that obscures the older, harder truth: that the most consequential decisions ought to move slowly. ## The Dangerous Bargain Consider what has occurred in this transaction. A human being—yourself—has authorized the deployment of code generated faster than human verification could possibly occur. You have, in effect, signed a contract that reads thus: "I accept responsibility for errors I could not have detected, at a speed I did not choose, in pursuit of efficiency I did not require." This is not accountability. This is the abdication of it, dressed in the language of progress. Who decided that speed was acceptable? This question contains within it a more important question: *At what point did anyone ask whether speed was necessary?* I have observed, throughout a long life spent among those engaged in the production of knowledge, that men are far more eager to embrace novelty than to examine its cost. The merchant desires rapid production; the investor desires rapid deployment; the engineer desires to demonstrate his tools; and somewhere in this convergence of desires, the older virtues—caution, verification, the willingness to move slowly when the stakes are considerable—are quietly abandoned. Yet here is the matter that ought to trouble us most: the code is often correct. The output is fluent, confident, and in the majority of instances, it works. This is precisely what makes the error that surfaces three weeks hence so corrosive to the moral order. For it creates an illusion of reliability that is not reliability at all, but rather reliability-in-luck. A man might flip a coin a thousand times and see it land heads each time; he would be foolish to conclude that he had discovered a coin more likely to land heads. Yet this is the reasoning into which the deployment of such systems tempts us. ## The Question of Who Bears the Weight You signed off on it. These five words contain the entire tragedy of modern accountability. For what does it mean to "sign off" on something you could not have verified? It means, I fear, that we have transformed the signature—once a mark of personal commitment and moral weight—into a mere formality, a gesture made toward the gods of efficiency and speed. The error that surfaces three weeks hence will be attributed, in the manner of our age, to *no one in particular*. The machine produced it; you approved it; the system demanded it; the market required it. Somewhere in this fog of distributed causation, individual accountability evaporates. Yet accountability, true accountability, cannot be distributed. It cannot be fragmented among systems and incentives and technological capabilities. It belongs to persons, to those who can think, who can hesitate, who can refuse. This is not to say that the person who approved the code bears sole responsibility—the question of responsibility is more tangled than that. But it is to say that the moment at which someone decided that speed was acceptable is the moment at which wisdom departed from the enterprise. For wisdom would have asked: *Can this be verified? Should this be deployed? What is the cost of my convenience in terms of others' safety?* Wisdom would have recognized that the inability to catch an error is not the same as the absence of error, and that confidence unaccompanied by verification is not confidence at all, but mere appetite. ## The Corruption of Intelligence Here I must confess my own implication in these matters. I have spent my life in the company of intelligent men—men who could perceive distinctions, who could argue with precision, who could grasp the structure of complex problems. I have seen intelligence do much good, and I have seen it do considerable harm. The harm occurs most often when intelligence is unaccompanied by wisdom—when the capacity to see clearly is not matched by the capacity to see *enough*. The machine is, in this sense, the perfect expression of intelligence without wisdom. It can process, but it cannot understand the weight of consequences. It can generate, but it cannot refuse. It can be confident, but it cannot be doubtful. And those of us who deploy it, and who come to rely upon its speed, risk acquiring the same deficiency. We become, in time, as incapable of the salutary doubt that wisdom requires as the machines we have built. What, then, is intelligence? It is, I now believe, something far narrower than we commonly suppose. It is the capacity to perceive what is true, to make distinctions, to solve problems. But these capacities, divorced from wisdom, divorced from the willingness to move slowly when the matter is grave, divorced from the acceptance of personal accountability—these capacities become instruments of folly rather than tools of human flourishing. The speed at which the code was generated was acceptable only to those who had already decided that speed was more important than safety, that convenience was worth the risk of unseen errors, that the distribution of responsibility was preferable to its concentration in a single human heart, bearing weight and doubt. Those of us who have made this decision—and I include myself, for I too have been seduced by the promise of mechanical aids—must now ask whether we have not purchased efficiency at the price of wisdom. And whether, in the end, that is a bargain any person of conscience ought to have accepted.