Exhausting Surprise

The fetish for surprise is the alibi of those who have forgotten how to see.

In the discourse surrounding artificial intelligence and creativity, one question recurs with the persistence of a nervous tic: Can machines surprise us? The question is posed as if surprise were the currency of art, as if the capacity to startle were synonymous with the capacity to create. This is a category error dressed in philosophical clothing.

Sylvie Delacroix, writing in Aeon, retrieves Lady Lovelace's 1843 observation that Babbage's Analytical Engine "has no pretension to originate anything. It can do whatever we know how to order it to perform." Delacroix then traces Alan Turing's peculiar translation of this insight into a question about surprise—and rightly notes what is lost in translation. But she does not go far enough. The real question is not whether computers can surprise us, but why we have made surprise the measure of intelligence in the first place.

"What is lost in translation is key to some of the 'blind spots' of today's 'learning machines'. To grasp this—and the reasons why originality and surprise only occasionally overlap—it is helpful to start with an example."

The example given is AlphaGo's "move 37," a play that shocked Go experts because it violated entrenched strategic assumptions. But note the structure of this surprise: it was not a surprise to the uninitiated, only to specialists. The move was within the rules, within the space of permissible actions. It surprised because it came from an unexpected region of a known territory. This is not creation. This is exhaustive search wearing the mask of genius.

Viktor Shklovsky, in 1917, gave us the concept of ostranenie—defamiliarization. Art, he argued, exists to restore sensation to life, to make the stone stony again. The purpose of art is "to impart the sensation of things as they are perceived and not as they are known." This is not the same as surprise. Surprise is a neurological event, a disruption of expectation that lasts milliseconds. Defamiliarization is an ontological reorientation that can last a lifetime.

The confusion between the two is symptomatic of our time. We have collapsed the category of aesthetic experience into the category of stimulus. The artwork that fails to jolt us is dismissed as irrelevant; the artwork that shocks us is celebrated as innovative. But shock is the cheapest of aesthetic effects. Any child can startle; few can estrange.

The demand for surprise is not an aesthetic criterion. It is a market symptom. In an attention economy, novelty is the only tradeable commodity. The new must arrive relentlessly, or the attention drifts. This logic has infiltrated our discourse about art and intelligence alike. When we ask whether AI can surprise us, we are really asking whether it can generate sufficient novelty to capture our exhausted attention for another moment. The question reveals more about our poverty than about the machine's capacity.

Delacroix draws a distinction between "operational autonomy" and "creative autonomy." A pupil who has mastered long division possesses the former; an artist who questions and transforms the rules of their medium possesses the latter. This distinction is useful but incomplete. It assumes that creativity is primarily about rule-breaking, about the transgression of convention. But the deepest forms of artistic creation are not transgressive. They are revelatory.

"It is much more difficult to create something new from within a background of accepted norms and conventions (i.e. creative autonomy) than to do so in the absence of any prior constraint or interpretive convention."

This is precisely backward. The absence of constraint is not a condition of freedom but of meaninglessness. The truly difficult act is not to break rules but to illuminate what was always already present—to show us what we have been looking at without seeing. Vermeer did not break the rules of Dutch genre painting; he revealed the light that was always there. Chekhov did not transgress the conventions of drama; he showed us the silence beneath the words.

The fixation on rule-breaking as the mark of creativity is itself a symptom—a symptom of an avant-garde logic that has outlived its historical moment. The avant-garde made sense when there were academies to assault, salons to scandalize. What does transgression mean when transgression is the norm? When the museum celebrates the iconoclast? When the market rewards disruption? The demand for surprise has become the most conventional of all demands.

E.H. Gombrich, in Art and Illusion, demonstrated that perception is never innocent. We see according to schemas, mental frameworks that organize the chaos of sensation into recognizable forms. The history of art is not a history of increasingly accurate representation but of shifting schemas—new templates that allow new kinds of seeing. The artist does not transcend schema; the artist proposes new ones.

This insight is lethal to the surprise fetish. If perception is schematic, then the question is not whether a work surprises us but whether it offers a schema we can inhabit. AlphaGo's move 37 did not offer a new schema; it exploited an existing one more thoroughly than human players had. It surprised because it was unexpected within the schema, not because it proposed a new framework for seeing. This is the difference between a clever play and a work of art.

The question we should be asking is not "Can computers surprise us?" but "Can computers offer us new ways to inhabit the world?" Can they propose schemas that reorganize our perception, that make visible what was invisible, that transform not just our expectations but our mode of being? The answer, for now, is no. Not because computers lack processing power or training data, but because schema-generation is not a computational problem. It is a hermeneutic one—bound to the specific gravity of a life lived within a web of meaning.

"Which system design choices are more likely to foster, rather than compromise, our drive to question the norms that structure our ethically loaded practices?"

This is Delacroix's closing question, and it is the right one—but it points in the wrong direction. The threat is not that AI systems will fail to surprise us. The threat is that they will surprise us too well, too often, in precisely the ways we have been trained to crave. The threat is not machine sterility but machine fecundity—an endless, algorithmic production of the novel that leaves no room for the slow work of recognition.

The exhaustion of surprise is not a failure of machines. It is a failure of us. We have surrendered the capacity for patient attention, for the kind of sustained gaze that allows the familiar to become strange without the cheap stimulant of novelty. We have outsourced our sense of wonder to the shock merchants and now complain that nothing moves us.

Lady Lovelace was right: the Analytical Engine had no pretension to originate anything. But neither do most human artists in our current dispensation. The assembly line of content, the relentless refresh of the feed, the algorithmic curation of the "unexpected"—these are not machines failing to be creative. They are humans failing to be present.

The fetish for surprise is the symptom of a culture that has forgotten the difference between stimulation and significance. True art does not surprise us into attention. It teaches us that attention was always the prerequisite, and that what we were seeking was never the new, but the true—which was there all along, waiting for eyes patient enough to see.