Artificial Intelligence Will Spare the Best but Replace the Rest

Just open the Events tab on social media, and a dazzling array of posters will fill your screen. Although each one invites you to a different event, they all look strikingly alike, as though produced by the same graphic design studio.

The same is true of thumbnails for podcasts and videos. Each promises something original, yet they resemble one another in telling ways. This similarity, too, is a product of the growing use of artificial intelligence.

The simplicity and accessibility of these tools have caused a genuine glut of generated images. A keen observer has thus learned to recognize a "work" produced by an AI image generator. Yet when we try to define precisely what gives it away, the task proves harder than it looks.

Of course, a practiced eye can spot errors in the details across a large number of images, such as the number of fingers on a depicted figure's hand. But these graphical hallucinations are something of a dead end in this search for an answer.

It is only a matter of time before these models improve further and such glitches become rarer still. High-quality versions already exist today, though only behind a paywall. The gap between generations of models is enormous, and such errors are diminishing fast. Basing criticism of AI-generated graphic design solely on these errors is therefore not particularly reasonable.

You might be interested How AI Is Transforming Weather Forecasts

The Hidden Logic Behind AI's Aesthetic Sameness

Something else is far more interesting. Even when we look at a set of AI-generated images with no obvious errors, we still cannot shake the feeling that they have something in common.

Most posters advertising events aim to draw people in with a positive image. A sunset is one of those aesthetic experiences that is not the exclusive privilege of the educated elite or of any particular social class. Its beauty could be admired just as readily by the German philosopher Immanuel Kant as by someone who has never once paused to consider what aesthetic experience even means.

Kant, in fact, was the one who tried to explain why our aesthetic judgments feel like more than merely private pleasures, why we sense that others might share them too. Artificial intelligence draws directly on this shared conception of beauty, distilled into the image of a setting sun. Pleasant pastel colors thus dominate the overwhelming majority of posters.

This homogenization of aesthetic output, however, is not a flaw in artificial intelligence but a fundamental principle of how it works. It can never be fully eliminated. It can only be concealed more effectively.

The AI model selects the statistically most likely visual combination, and to create a positive effect, it typically reaches for warm colors and soft lighting. As newly generated images feed back into the training data, moreover, this tendency risks becoming even more pronounced.

Artificial intelligence thus constantly reproduces and reinforces the same pattern, pulling everything toward mediocrity, which, on an aesthetic level, produces kitsch in the style of Walt Disney. His cartoon characters, with their big eyes, are sweet and saccharine, which lets them appeal to a broad audience. Yet all of his fairy tales end up blending into one, bearing only the Walt Disney brand.

You might be interested AI Learns to Dream in the Race for Smarter Machines

The Illusion of Originality, from Pop Songs to AI Art

The problem of originality in AI-generated images is nothing new. The philosopher Theodor Adorno addressed it in detail decades ago, labeling it "pseudo-individualization".

Adorno did not develop this theory from visual production. Instead, he analyzed the popular music of his time. The music industry, he argued, produces songs from the same template, varying only in minor details, so that listeners get the impression they are hearing something original. In truth, nothing exceptional is taking place. Individuality exists only on the surface. Adorno added, however, that this pseudo-individualization is essential, since without it audiences would quickly lose interest.

AI fulfills this requirement perfectly. It has no difficulty producing infinite variability without any true originality. Where it will struggle is with originality of the human kind. Genuine originality typically requires breaking the mold, whereas a generative model, by its very nature, cannot stray from the patterns it has already learned.

That said, people should not celebrate too soon, because real originality is exceptionally rare to begin with. Contemporary art makes the point well. Amid the flood of human creative output, very little today is genuinely original. This is not a criticism of the current state of affairs but simply a description of how things work. Originality is as scarce as saffron, and saffron would not be so prized if it were not scarce. On this front, artificial intelligence will struggle to compete.

You might be interested Frustrated Young Couples Turn to AI Romantic Partners

AI's Target: Everything Between Genius and Kitsch

Does this mean creative professionals have nothing to fear for their jobs? Not quite, but the real threat lies elsewhere. The purpose of pseudo-individualization is, above all, to create the impression that a work is not interchangeable, and this is precisely where AI can compete with established creative industry.

Most customers do not need work that looks as though it came from Picasso or Dalí. It is enough if their visual design is not interchangeable with a competitor's. Only a handful of companies require the kind of originality that goes on to shape the history of modern design.

This distinction extends well beyond graphic design. Much the same dynamic is likely to play out in music, writing, video and animation. In each of these fields, a large part of the market does not demand genuine originality, only competent work that feels sufficiently distinctive. AI does not need to write the next great novel, song or film to transform these industries. It merely needs to become good enough at producing the broad middle ground of professional creative work.

Greater integration of artificial intelligence into creative work is inevitable. It will not replace true originality, of course, and the price of that originality should rise. People will still have something to offer.

The problem, however, will not lie on the supply side but on the demand side. An ever-smaller portion of the market will be willing to pay for originality, not because of its cost, but because they will have little practical use for it.

Artificial intelligence will therefore most likely take over much of the vast middle ground of creative work, from graphic design to music, writing and beyond: everything that falls between genius and kitsch.