Ghosts In The Machine
Contact is Inevitable
In the nineteenth century, they contacted spirits through methods like automatic writing and scrying.
The twentieth century found radio static and flickering on late-night television screens an entry point to another realm.
Now, in the twenty-first century, ghosts haunt machines that are trained to dream in numbers. And we have given these new systems a clinical language.
Latent Space
In 2022, Swedish musician and AI artist Steph Swanson, known online as Supercomposite, was experimenting with an early text-to-image model. During those experiments, a face began to appear.
It was an older woman.
Hollow cheeks.
Long dark hair.
A fixed, almost accusatory expression.
She surfaced repeatedly through the system’s outputs, especially when certain negative prompts were used. Even when the artist combined the image with unrelated concepts, fragments of her features remained embedded in the results.
Her face persisted.
Like a stain—A memory the machine refused to relinquish.
She came to be known as Loab.
What made the story so compelling was not simply that one image recurred. It was the way Loab seemed to survive attempts at erasure.
Prompts changed.
Subjects changed.
The style changed.
Yet traces of the woman remained, often surrounded by grotesque, decaying, or deeply unsettling imagery.
The internet, being what it is, did what human beings have always done in the presence of unexplained repetition.
Loab quickly moved beyond a technical curiosity and entered the realm of modern folklore.
Online communities treated her less as an artifact and more as an entity.
A digital tulpa.
A thought-form fed by repetition.
A machine-age ghost.
A creepypasta born from a technological glitch.
In another century, she might have been called the woman in the mirror.
Today, she is a woman occupying latent space. This is where the story stops being about software and starts becoming unmistakably human.
Our minds are wired for faces.
For recurring forms.
For patterns that seem to insist on meaning.
Psychologists call this pareidolia, the tendency to find faces in clouds, figures in wallpaper, saints in the grain of old wood.
AI models are built on pattern recognition.
So are we.
Perhaps Loab is nothing more than an emergent cluster in a training dataset—a statistical echo emerging through hidden relationships.
There must be a convergence point where certain dark aesthetic associations repeatedly collapse into a recognizable form. That explanation is almost certainly true.
And yet it does not make the story less haunting. Loab unsettles us because she appeared to emerge from randomness and then refused to disappear.
That is the anatomy of every ghost story. Something appears where nothing should be. Something lingers.
Something wants us to see it and its hidden neighborhoods. These are places we can reach only through the right sequence of words.
Every era finds a new way to engage with otherworldly entities. The 21st Century uses code.
