AI uncovers no mystery of art


People tend to rejoice in the disclosure of a secret.

Or, at the very least, the media has come to realize that news of “mysteries solved” and “hidden gems revealed” generates traffic and clicks.

So I’m never surprised to see AI-assisted revelations of famous masters’ artwork go viral.

In the past year alone, I have come across articles highlighting how artificial intelligence has recovered a “secret” painting of a “lost lover” by Italian painter Modigliani, “brought back to life” a “hidden naked Picasso “,” Resuscitated “an Austrian painter The destroyed works of Gustav Klimt and the” restored “parts of Rembrandt’s 1642 painting” The Night Watch “. The list is long.

As an art historian, I am increasingly concerned with the coverage and circulation of these projects.

They haven’t, in fact, revealed a secret or solved a single mystery.

What they’ve done is generate AI wellness stories.

Are we really learning something new?

Take the reports on the paintings of Modigliani and Picasso.

These were projects carried out by the same company, Oxia Palus, which was founded not by art historians but by machine learning doctoral students.

In both cases, Oxia Palus relied on traditional X-rays, X-ray fluorescence, and infrared imaging that had already been done and published years before – work that had revealed preliminary paintings under the visible layer on them. canvases by artists.

The company edited these x-rays and reconstructed them as new works of art using a technique called “neural style transfer.” This is a fancy-sounding term for a program that breaks down works of art into extremely small units, extrapolates a style from them, and then promises to recreate images of other content in that same style.

Essentially, Oxia Palus assembles new works from what the machine can learn from existing x-ray images and other paintings by the same artist.

But aside from flexing the prowess of AI, is there any value – artistically, historically – in what the company does?

These recreations teach us nothing that we did not know about the artists and their methods.

Artists paint over their works all the time. It’s so common that art historians and restorers have one word for it: repent. None of these earlier compositions was an Easter egg deposited in the painting for later researchers to find out. The original x-ray images were certainly valuable in that they offered insight into the artists’ working methods.

But for me, what these programs do is not really interesting from an art history perspective.

The human sciences on the support of life

So when I see these reproductions catching the attention of the media, it strikes me as soft diplomacy for AI, showcasing a ‘cultured’ application of technology at a time when skepticism about its deceptions, prejudices, and ideas. abuse is on the rise.

When AI calls attention to the recovery of lost artwork, the technology looks a lot less frightening than when it makes headlines for creating deep fakes that distort politicians’ rhetoric or for using technology. facial recognition for authoritarian surveillance purposes.

These studies and projects also seem to promote the idea that computer scientists are better at historical research than art historians.

For years, university humanities departments have been steadily deprived of funding, and more money has been spent on the sciences. With its claims to objectivity and empirically provable results, the sciences tend to command greater respect from funding agencies and the public, prompting humanities researchers to adopt computational methods.

Art historian Claire Bishop has criticized this development, noting that when computer science becomes part of the humanities, “[t]theoretical problems are overwhelmed by the weight of data ”, which generates deeply simplistic results.

Basically, art historians study the ways in which art can offer insight into how people once viewed the world. They explore how works of art have shaped the worlds in which they were created and will influence future generations.

A computer algorithm cannot perform these functions.

However, some academics and institutions have let themselves be subsumed by the sciences, adopting their methods and partnering with them in sponsored projects.

Literary critic Barbara Herrnstein Smith has warned against giving too much ground to science. In his opinion, the sciences and the letters are not the antipodes that they are often presented publicly. But this representation benefited the sciences, valued for their supposed clarity and usefulness over the supposed obscurity and uselessness of the humanities. At the same time, she suggested that hybrid fields of study that merge the arts and sciences could lead to breakthroughs that would not have been possible if each had existed as a siled discipline.

I am skeptical. Not because I doubt the usefulness of expanding and diversifying our toolbox; Certainly, some researchers working in the digital humanities have adopted computational methods with subtlety and historical awareness to qualify or reverse entrenched narratives.

But my lingering distrust emerges from an awareness of how public support for the sciences and denigration of the humanities means that, in the effort to gain funding and acceptance, the humanities will lose what makes them. vital. The domain’s sensitivity to historical peculiarities and cultural differences makes the application of the same code to a wide variety of artefacts totally illogical.

How absurd it is to think that black and white photographs from 100 years ago would produce color the same way digital photographs do today. And yet, that’s exactly what AI-assisted colorization does.

This particular example may seem like a small qualm, of course. But this effort to “bring events to life” systematically confuses representations with reality. Adding color does not show things as they were, but recreates what is already a recreation – a photograph – in our image, now labeled by IT.

Art as a toy in the scientists’ sandbox

Towards the conclusion of a recent article devoted to the use of AI to disentangle the radiographic images of the “Ghent altarpiece” by Jan and Hubert van Eyck, the mathematicians and engineers who wrote it refer to their method as relying on “choosing” the best of all possible worlds (borrowing Voltaire’s words) by taking the first exit from two separate tracks, differing only in the order of the entries.

Perhaps if they had familiarized themselves more with the humanities, they would know how satirical those words were when Voltaire used them to make fun of a philosopher who believed that endemic suffering and injustice were all pervasive. part of God’s plan – that the world as represented the best one could hope for.

Maybe this “gotcha” is cheap. But it illustrates the problem of art and history becoming toys in the sandboxes of scientists with no humanities training.

At the very least, I hope the journalists and critics reporting on these developments take a more skeptical eye on them and change their framing.

In my opinion, rather than seeing these studies as heroic achievements, those charged with reporting their findings to the public should see them as opportunities to question what computer science does when it takes ownership of the study. art. And they should ask themselves if this is all for the good of someone or anything other than AI, its most zealous supporters and those who profit from it.

This article from Sonja Drimmer, Associate Professor of Medieval Art, University of Massachusetts Amherst is reposted from The Conversation under a Creative Commons license. Read the original article.


Comments are closed.