“This will be the year of AI,” one of my colleagues at Virginia Tech proclaimed rather grandly as we assembled for the new school year a few weeks ago. “And I am not sure we are equipped for it.” What he meant was that a host of new artificial intelligence-driven software programs and apps, most prominently Midjourney and Dall-E, are now so widely available that our students will be—and, as became evident very quickly, already are—producing images with a lushness, range of references, and sophistication that indeed threaten some of the basic tenets of teaching. What does it mean that students can create images in a few hours that on the surface are so much more accomplished than anything their teachers have ever made? Is it even their work or that of an algorithm? How do we judge what we see?
In some ways, the debate is an extension of the discussions that ensued way back in the First Silicon Age of the early 1990s (akin to the Stone Age in modern time reckoning) when students began unleashing their blobs and meshes on virtual drafting tables, often with far more skill than their fuddy-duddy professors. In other ways, though, something new is indeed here for us to reckon with: The images produced by Midjourney and Dall-E have a degree of realism, a range of painterly effects, and a way of plugging into the history and traditions of architecture that cloaks their novel mode of production with disarming familiarity. The images also often beat those arguing for hand-drawing and modeling techniques at their own game. The work looks more hand-made, more realistic and more crafted, more knowing in its references, and grander in its ambitions than the collection of swoops and swerves we have come to understand as the marks of computer technology.
To get a sense of the possibilities, you need only turn to the recent work of the Madrid-based Italian architect Cesare Battelli. A former collaborator of Enric Miralles, who was trained at Frankfurt’s Städelschule when that school was a hotbed of both technological and image innovation under Peter Cook, he has for the last decade concentrated on being a “visionary architect,” as he says. Six months ago, he began using Midjourney and posting some of the results on social media. They are some of the most sophisticated architectural images I have seen in the last few years.
The best of them, like the series that ensued from Battelli prompting the software to render the “Tower of Babel under construction,” show a world in which scaffolding and the fragments of finished forms wrap around each other to produce buildings that change from solid objects to ephemera as they rise up into the sky. The scenes Battelli produces almost always consist of such unfinished structures, all loosely connected and spreading out across the virtual page in a sepia-tinted continuum. Small and indistinct figures toil away at the structures or admire them from below. Bits and pieces of machines push out of parts of the buildings.
In other series, where Battelli combined prompts for a particular building type with animals such as sea horses or tigers, curved covers that might be roofs or walls turn into shapes that evoke the fauna before your eyes—all without ever being distinct enough to be completely recognizable. Diving into Battelli’s images, you concentrate on elements you think you know, from construction pieces to walls to hides, only to watch them melt into something else before you can focus on the particular properties of a part.
What binds these fragments together is not only how the algorithms do not respect the finished and distinct aspect of any of the inputs, instead continually developing or fading them together but also the atmospherics (a kind of romantic sauce) that makes Midjourney both more fairy tale in its appearance and more sophisticated than most of the other readily available software. Battelli incorporates phrases such as “in the manner of” and “in the style of” in his prompts and has a penchant for a particular range of early Renaissance Italian painters, including Giotto and Botticelli. He adds the hazy skies and attenuated forms of Tintoretto and Veronese to the mix but also calls on compendiums of scientific data produced during that period as reference material. These references give the images their sense of emerging out of a Western art history tradition that removes the science fiction or romance novel aspect that overwhelms most of the other Midjourney work.
Battelli is particularly fond of the obscure artist Monsù Desiderio, a painter working in Naples during the early 17th century. Virtually nothing about him is known, and the name might even indicate the work of three different artists originally from Germany and France. “He is a ghost,” Battelli says, “whose work is also spectral.” A proto-surreal collection of ruins and construction sites, explosions, fires, and refined structures, Desiderio’s images indeed presage the collage-like sensibility that AI programs generate.
So what is it that Battelli does? He uses a palette that consists not of paint or pixels but rather of references. He is becoming increasingly skilled at choosing sources to create the effects he wants. It is, he says, a way of “alchemically creating something unexpected,” made possible because he works “on the threshold” where images and ideas solidify. He calls the result a product of imagination, distinguishing that evocation from utopian or dystopian visions because it works on and with existing images and forms. It is thus an advanced mode of collage that takes pre-formed materials and assembles them into a snapshot of intensified relations and compacted juxtaposition. Through the unifying tendencies of Midjourney, that collage, which is the Western tradition would usually remain a collection of distinct elements, is smoothed into something that denies its own fragmentary origins.
Recently Battelli has also been able to have Midjourney produce what appear to be wood models of some of his drawings. Now he is imagining that he might take these artifacts and continue them in balsa wood and glue, extending what currently exists only on the screen into the three-dimensional world. The models might then, he hopes, become part of a collage we can inhabit in the mode of an art installation. I can easily imagine one of my students taking a photograph or a video of that work and asking Midjourney to produce an image of their thesis project in the style of Cesare Battelli. Or an architect asking his computer to produce an image of the office building he is designing in the style of Battelli channeling Desiderio. The possibilities are endless, and, I for one, am pretty excited by the extension of architecture that artificial intelligence is opening.
This article has been updated.
The views and conclusions from this author are not necessarily those of ARCHITECT magazine or of The American Institute of Architects.
Read more: The latest from Aaron Betsky includes walking the streets of the Postmodern "theme park" that is Portmeirion, exploring Assemble's neighborhood activism in Liverpool, and weighing in on an addition to a Neo-Brutalist monument in Boston.