To make a virtue out of crudeness, there seem to be two attitudes that one could take towards artificial intelligence. Either one is moved to wonder and excitement, or one is more mistrustful, cognisant of the various ways in which AI might alter—corrupt—the ways we live and interact with each other. But whatever stance one leans towards, the unstated assumption is that AI’s impact will be primarily sociological rather than existential. The debate invariably centres on the ways in which the technology could be used—or misused. I wish to suggest, however, that AI really ought to evoke uneasiness of a more elemental pedigree, that strikes at the heart of—to invoke a colossal cliché—what it means to be human.
We are seldom troubled that AI can already do many things much better than us. We welcome the computer’s superhuman calculation abilities. We might even stomach, if grudgingly, that no human now stands a chance against an engine in chess, go, or shogi. With the recent developments in machine learning and large language models, it is increasingly difficult to claim that humans are more ‘intelligent’ than computers in any substantive sense. But still, they are just programs. The prowess of the invention redounds to the glory of the inventor. The unexceptionable belief in our special place in the world—the default and only position we have ever known—seems as secure as ever.
For many, the conclusive proof of our superiority lies in our consciousness. Machines may simulate emotions, but they can never feel them. A program lacks a mind, so while it can ‘generate’ artworks or poetry by imitation and collage, these works can never match the genius of human masters. Yet to me, it is precisely this kind of reasoning which suggests a Copernican revolution may be at hand. The question is: are our criteria for ‘sensibility’ and ‘consciousness’ robust enough to cement forever our dominance over the machines? Imagine that one day there will be programs which could generate works of art and literature splendid enough to eclipse Shakespeare or Rembrandt—what sort of world would that be? We may accept humanity’s defeat in chess with only a moderate shock, but art is a precinct too close to the heart of humanity. The existential doubt might become too potent to fob off: what exactly, then, are we good for?
In ancient China, the game of Go was ranked among the four aristocratic arts, alongside calligraphy, painting, and the string instrument guqin. To play Go was more than entertainment: it was simultaneously an artistic exertion and a metaphysical foray. Until the 2016 match between Lee Sedol—top Go player and winner of eighteen international titles—and the program Alphago, many thought that Go would remain a bastion against the machines, a quintessential example where refined sensibility still triumphs over brute-force calculations. The match would upend the Go world—Lee lost 1-4, and the game he won is now enshrined as a miracle. The theories of Go that perdured through centuries now had to be rewritten: computer programs had learnt Go and reached such a height of mastery that their moves are, more often than not, no longer comprehensible to even the best human players.
But perhaps the same fate will not befall the world of art. Go is ultimately a competitive game, one is tempted to say, where the value of a move could be determined objectively, but art is fundamentally subjective. This is no doubt true, but unless one is willing to concede that art is no more than what the art-world declares to be art—a view that the philosopher Arthur Danto held—there must be something in the judgment of an artwork which is, if not “objective”, at least amenable to reason. If I say that Hamlet is a masterpiece, this cannot merely be the expression of an arbitrary preference: there are, so to speak, evidence that I could adduce, textual details that, through reasonable interpretation, bring to life the playwright’s genius and vision. In other words, there must be things that I could point to and say, can’t you see how great it is? In saying so, I ask my interlocutor to test the strength of my evidence against their sensibility. It is never guaranteed that they agree—and to this extent aesthetic judgements are not objective but intersubjective—yet there must be a baseline of rationality which renders my evaluations more than the assertion of tastes. Perhaps this is enough of a foothold for AI.
The claim that computers cannot create great artworks because they lack a mind rests on a flawed understanding of intentionality. When one examines an artwork closely and asks, what is the artist’s intention? one is reaching for a plausible hypothesis and not data of the artist’s psychological state. We speak of a director’s intention in a movie scene, but could the actors not have decided it on a whim? In assessing the value of an artwork such a question is irrelevant. The ‘artistic intention’ is really an explanatory device—we would not accept an artist’s testimony unless we could find it confirmed in the artwork. In other words, the connection between an artwork’s rich interpretive possibilities and its conception by a mind may only be a contingent one. To put more pressure on the idea, we might imagine an advanced AI algorithm proceeding much as the human artist: it might generate a rough blueprint, evaluate it, fill in the details as it goes, pausing for evaluation, back-pedalling when necessary for revision. If it is indeed coherent to speak of something ‘making an artwork great’, is it so inconceivable that a program, having digested terabytes of aesthetic data, might be able to produce that something in more copious amounts than anyone before?
Arguably, the situation here may be more dire than what happened with chess or Go. Just as cars do not pose a threat to athletes, chess and go engines are not rivals of human players. In art, however, machines and humans may be in direct competition. If we were to discover that Shakespeare’s plays were actually generated by machines, we would not thereby strip them from the literary canon. If the day comes when AI art would exhibit greater aesthetic value—by whatever standard we choose—than any artwork created by humans, there is no reason why we should not embrace the development. After all, is the primary reason we value art not that it moves us with its beauty and depth? This comes, of course, with the chilling thought that one day we might listen to Beethoven’s symphonies with interests that are more anthropological than aesthetic—like how we regard prehistoric wall paintings.
And here is the other sense in which the possibility of AI art touches a neuralgic point: art has long been the altar on which artists deposited their dreams for immortality. For Sartre, it is the final site of the ‘impossible synthesis’ where one seeks to suspend one’s finitude in a dream for transcendence. To be forgotten is sometimes more frightening than death, and immortality through art is, if an illusion, at least a comforting one—but now AI art threatens to disrobe it. Maybe we are entitled to hope, to take heart from the fact that the blistering developments of modern art have not obsoleted Rubens, but perhaps an existential reckoning is already inescapable.