AI Meets the Romantic in Ned Richardson’s Show at the Front

The rise of generative AI over the past three years has brought with it a downward spiral of plagiarism and misinformation, environmental costs, and job losses. Less important, but no less sad, is that now it’s finally here, the technology imagined by countless sci-fi writers as a new consciousness, capable of dreaming of electric sheep, has instead manifested itself in the most boring way possible: as an entity that gives us exactly what we ask for.

But when it doesn’t — when AI delivers invented results known as “hallucinations” — things start to get interesting. In such instances, language allows our conception of technology to slide toward poetry, as it does in Ned Richardson‘s solo exhibition “new memory: landscapes, ghosts, machines,” on view through June 1 at the Front in Montpelier.

Richardson, 65, of Moretown, started building his own generative adversarial networks around 2018, years before AI technologies such as ChatGPT were publicly available. He fed images of his own landscape paintings to the GAN. The model works through two neural networks: The generator creates new pictures based on the inputs, while the discriminator determines whether an image is real (from the input dataset) or a fake (generated by the network).

The GAN images are square and pixelated; squinting at one, we seem to see a landscape, with billowing black forms like trees, blue-green expanses, cloudy blobs. Richardson has incorporated them into many of the works in the gallery.

“rrun : 0110 or 1001” consists of four eight-foot panels, each with eight of these GAN images on it; transparencies that look like filmstrips run below each set, with dozens of thumbnails of similar images. Some look like static or noise. “That’s what happens when the training collapses,” Richardson said on a tour of the show. Some of the images have a pinkish cast, an attempt to introduce flesh tones to the set to approximate the figure in a landscape.

From early 2022, “rrun : 0110 or 1001” is a kind of base code for the rest of the show. It brings in layers of meaning, text and reference while showing the images in their original state, without any of the manipulations present in other works in the show. The “rrun” in the piece’s title refers to “riverrun” in the first line of James Joyce’s Finnegans Wake. Each image is part of a 1 or 0 set, which combine to make letters based on ASCII encoding. Spelled out, those codes give each panel its own subtitle, and the subtitles together make an appropriately Joycean kind of poem: “bild es kin / all with in / natùr hem / gone go lem.”

None of this would be obvious from just looking at the work, and it’s not even necessarily important. But it is pretty representative of how Richardson thinks, pulling in bits of literature and math, playing linguistically with programming terms, making new monsters out of spare parts — he also cites Mary Shelley’s Frankenstein as an important influence.

Coming out of the pandemic, Richardson found himself reconsidering his work as large language model AIs such as ChatGPT suddenly became commonplace. He made a series of 8-by-8-inch paintings, all called “untitled element,” that incorporate GAN images with materials such as cold wax, pastels and glass. “untitled element [6] [Frankenstein’s apple]” includes a painted still life of an apple, a film strip and layers of paper from a wasp’s nest. “untitled element 4” is entirely black, the GAN image buried under a waxy, tar-like substance that’s been etched and scratched into.

“I really internalized the imagery, and I really connected intuitively with the process — the idea of training a model,” Richardson said. “We don’t live in the world. We live in a model of the world that our brain synthesizes from our visual input and our auditory input. That’s what we inhabit, is the model that our brain makes.”

Many of the works in the show riff on that idea of modeling, playing more with the idea of how we see than with what we see. Several layer GAN imagery printed on transparencies with parts from laptop and phone screens and pieces of glass onto which Richardson has etched drawings. “untitled [ghost landscape 2]” floats an image over a piece of wood — an artifact from the real landscape. In “untitled [shore],” an etching hovers over a mirrored surface taken from inside a tablet.

Some of these works are made up of translucent layers and literally backlit when taken off the wall. They change dramatically with the light in the gallery. “untitled [logscale glass] [shipwreck, sunset]” incorporates woven, painted paper and an assembly including the surface from a digital touch screen, positioned as though the viewer were inside the machine, looking out.

Richardson has gone a step further with a series of glass stereograms, accompanied by vintage stereoscopes. The images were taken from a family collection dating back to the turn of the 20th century and picturing India, China, Egypt and Japan. Richardson deleted information from the scanned originals, creating ghostly transparencies sandwiched between etched glass surfaces. They almost resolve into 3D scenes when seen through the viewer, but not quite. The viewer is left with a sense of loss and nostalgia, augmented by the old-book smell of the stereoscope.

Ghosts recur throughout the show, from the idea of the “ghost in the machine” to ghosts of landscapes to ghosts as history. The theme is of a piece with the romantic sensibility pervading Richardson’s work, from ships and dewy landscapes to the Victorian “opium smoker” in one of the stereogram images. Such elements contrast with the work’s tech sensibility without contradicting it.

Richardson talked about the circular conception of history that informed Finnegans Wake, which ends where it started. Our current moment, he said, often feels like the Gilded Age, when stereograms were the cool new imaging tech.

“Circular” may not be quite the right descriptor, though, for how Richardson’s motifs and themes pop up in one place and come back around in another. In programming, it’s called recursion: going further and further into a spiraling function until eventually you’re back outside it, ready for the next tangent.