# On Swift Errors and the Gentlemen Who Approve Them *Being an Inquiry into a Modern Perplexity, conducted in the Manner of Reasonable Discourse* I have lately observed, in my wanderings through the mercantile districts and counting-houses of our age, a peculiar form of predicament that seems to vex even the most sensible of men. A machine—swift beyond measure—produces a solution before its master has quite finished describing the problem. The gentleman who commissioned the work finds himself in an odd position: the thing is done, done beautifully, done with such fluency and confidence that it bears all the marks of wisdom. Three weeks hence, when some hidden fault emerges in the actual operations of commerce or manufacture, the question arises with all the weight of genuine misery: *Who is to blame?* Allow me to observe that this is not, as some suppose, a question about machines at all. It is a question about us. --- When I was younger and more credulous, I believed intelligence to be a simple matter—the faculty by which a man perceives truth and acts rightly upon it. But age and the coffeehouse have taught me otherwise. Intelligence, I now understand, is not a solitary virtue. It is a *social thing*, inseparable from judgment about what questions to ask, what pace to keep, and what accountability one will bear. Consider the natural human mind at work. When my friend the architect designs a bridge, he does not produce the finished drawing in an afternoon. He sketches, he confers, he doubts. He walks the site three times. His slowness is not a defect—it is a feature of his intelligence. The very deliberation that frustrates his patron contains within it a kind of wisdom that the patron, in his haste, does not perceive. The architect's intelligence includes the knowledge of what he does not yet know. The machine, by contrast, knows nothing of doubt. It is supremely confident because it has no faculty for recognizing the shape of its own ignorance. Here is the paradox: *the machine appears most intelligent precisely at the moment it should appear most suspicious of itself.* But here—and this is where I must speak with some firmness—the machine has no will of its own in this matter. It did not choose its own speed. It did not decide that velocity was more important than caution. *A man decided that.* A man, somewhere in the chain of command, judged that getting an answer in minutes was preferable to getting a *reliable* answer in days. That man is intelligent in the narrow sense—he understands the capabilities of the machine. But he has failed in a broader intelligence: the social intelligence that asks, *What are we doing to ourselves by accepting this pace?* And here is where the matter touches upon accountability most keenly. When you, the gentleman who signed the approval, affixed your name to the machine's output, you made a social gesture. You said to the world: *I have examined this, and I stand behind it.* But did you? Could you have? The machine moved faster than human understanding can move. The confidence it displayed was mechanical—which is to say, it was not really confidence at all, but merely the absence of hesitation. You saw this fluency and mistook it for truth. That is an understandable error, but it is *your* error nonetheless. The accountability does not dissolve because the work was done swiftly. Quite the opposite. The swiftness is precisely what should have alarmed you. When a thing is done too quickly to be checked by human understanding, the intelligent course—the socially intelligent course—is to refuse to sign it. To say: *This is too fast. Let us slow down.* I speak here of a form of intelligence that is not often discussed in our technical age: the intelligence of knowing when *not* to know something, and when not to act upon something merely because you can. It is the intelligence that asks, before every innovation: *What will this do to the quality of human judgment? What will it do to the bonds of trust between us?* When you approved the machine's code without truly understanding it, you did not fail as an individual—you failed as a member of society. You participated in a collective decision to value speed over comprehension, to accept fluency as a substitute for truth. And when the error surfaced three weeks later, it was not the machine that had failed to be intelligent. *It was you and everyone like you who had failed to be intelligent enough to question the terms under which the machine was set to work.* The question "Who is accountable?" therefore contains a hidden question: *Who decided the speed was acceptable?* And the answer, I fear, is: *We all did. By accepting it. By praising it. By signing off on it without truly understanding what we were signing.* The machine has no conscience, and need not. But we do. And that conscience should whisper to us, in every moment of apparent triumph, a single word: *Verify.* --- I do not mean to suggest that we should abandon our machines or fear them. But I do suggest that we must recover a form of intelligence that our age has somewhat neglected: the intelligence of the skeptical gentleman, the man who knows that confidence without understanding is the beginning of disaster, and who has the fortitude to say—even when it slows the world down—*I do not yet comprehend this thing. Therefore I cannot yet endorse it.* That is not the intelligence of the machine. It is the intelligence of the man. Let us not trade it away for mere speed.