“A Masterpiece From Your Photos.” This is the promise made by AI Gahaku, one of the most popular artificial intelligence engines trained to generate images that resemble old paintings using our own pictures. It is simple as it sounds: You upload a photo of a face, choose the painting style, and voilà, the AI generates an image that looks like a century-old Western portrait ready to be displayed in a museum. The success of AI Gahaku follows the success of many other apps and projects that over the past few years have experimented with user pictures and AI filters. AI Gahaku claims that “various painting styles can be easily applied to it such as Renaissance, Pop Art, Expressionism and many more!”
It would be too easy to deem the popularity of AI-generated portraits resembling old paintings as due to the nostalgic enjoyment of figurative art styles that are safer and easier to understand than modern art. One should instead look at the questions raised in the process of generating images with an AI. When we upload a picture of ourselves and wait for the machine to finish its calculations, what do we expect to find in the image it generates? Why does the AI almost always fail at being lively or authentic?
An initial answer can be found in the socioeconomic developments that formulated the main aspects of 19th-century realist portraiture, which are similar to the developments that inspire contemporary artistic AIs.
In 1830, “The bourgeoise is in full possession and awareness of its power;” and “the aristocracy has vanished from the scene of historical events and leads a purely private existence,” according to Arnold Hauser, writing in his book The Social History of Art. The middle class plays a central role in the invention of new art forms, such as the Naturalist novel, and realist reinterpretation of genres that were historically connected to the display of aristocratic power. The desire to pursue factual reality in art mirrored the desire of the middle class to highlight values that made it a very different social group in comparison to the nobility which saw its influence on European culture in general fade in direct proportion to the rising influence of the bourgeoisie. While the aristocracy’s power and role in society were hereditary, not depending on the personal characteristics of its members, the middle class was interested in the qualities that made artisans and merchants powerful figures in society, that is, their personal character and self-determination.
The interest in being portrayed and remembered as unique figures led to a sprawl of commissions for individual and group portraits. The imagery of pictorial Realism was full of what might strike us as meaningless details, yet this emphasis on detail “ is what nails its productions down so firmly to a specific time and a specific place,” as American art historian Linda Nochlin explains in her book Realism (1973) “It anchors Realist works to a concrete rather than an ideal or poetic reality.”
For example, when we look at this 1861 portrait by French painter Émile Auguste Hublin, we have visual clues that allow us to provisionally determine whether the sitter is a jolly spirit, or a careful mother. We are tempted to wonder about her thoughts. What’s portrayed is not only the physical features but also the psychological dimensions of the individuals. If the 19th-century artists who painted realist portraits were asked to play this mediating, almost magical role, how could such a difficult creative task should be carried out well by an AI? In short, it cannot.
The sensibility of the modern art-viewing public is not the same as that of 19th-century bourgeoisie, but we still fall for the idealized and nostalgic aspects of their images, full as they are of elegantly dressed men and women. If artists made portraits in that Realist style now, their approach would read as kitsch. So says the Italian art historian Gillo Dorfles in the book Kitsch. The World of Bad Taste: “The people who made these camouflages of authentic avant-garde art have isolated one single aspect of the artistic phenomenon they try to imitate and which originally had a genuine creative value, raising it to the level of a pattern but thus depriving it of any novelty.”
Developers who train AIs to analyze and replicate recurring visual qualities of 19th-century bourgeois portraits imitate that style, isolating its pictorial qualities from what made it so relevant at the time, namely the social environment where it developed and became popular. Essentially, they copy the style, but lack the crucial substance.
In order to reproduce the 19th-century realist portrait, an AI should be trained not just to paint like artists of the time, but foremost, to think like them. We, the contemporary art-viewing public, do not find beautiful or meaningful these AI-generated faux old portraits in the same way the originals were enjoyed and appreciated by its contemporaries, though they inspire curiosity. They inspire us to be curious because they create unexpected versions of our appearance. When I upload a picture of myself in the AI app, I can’t know what features of my face will grab the attention of the machine and how these will turn out once they have been filtered through the algorithm. What these AI engines lack in creativity is made up by the genuine expectation of the viewer.
We are curious about how we would have looked like in a 19th-century portrait forgetting that only very few had the economic means and political power needed to be immortalized in a painting. Most of the bourgeois portraits made at the time show a very select group of sitters and weren’t meant to be on display in public galleries and museums. It’s our contemporary sensibility and approach to the art of the past — made so widely available thanks the internet — that lets us look at the portraits of that societal group as just another visual style that doesn’t carry any political value. We miss so much when we think of this imagery as just one filter among many others to choose from when we upload our picture into an AI app.