My AI Image Experiment In Dream Analysis
We've always expressed our nightmares through images. So one Italian writer fed her dreams to AI-powered Midjourney platform, producing images of her own consciousness.

Dream of April 13th.
TURIN — I have been writing down my nightmares for as long as I can remember: they are the starting point for my writing, and doing so is essential for my relationship to myself.
I am certainly not the only one writing these down: the transcription of dreams is one of the oldest literary genres. The first meticulous dream collection by the Englishwoman Anna Kingsford, a hardened animalist, dates back to the 19th century. I wonder if this is a coincidence. Or are those who pay attention to the subtle language of dreams also lovers of the animal kingdom?
The earliest nightmare ever transcribed, however, dates back as far as the third millennium B.C., and is found in the Sumerian poem The Descent of Inanna into the Underworld. It's the story of a descent into hell that strips the ancient goddess of love and war Inanna of one garment at a time until she is left naked in the presence of her monstrous sister Ereshkigal.
During this journey, which in itself already resembles a dream (the theme of nudity/insecurity has no doubt appear into many people’s dream world at one time or another), there is a nightmare in which the demons of the Underworld pursue Dumuzi — the god of shepherds, and fertility and consort of Inanna — and finally succeed in capturing him.
This is followed by the interpretation of her sister, a true professional dream interpreter: with the icy detachment of a true psychoanalyst, she confirms that the dream hides premonitions of death. Finally, an illustration: a series of seals representing the motifs of the dream.
In short, along with being the first transcribed dream in history, it's also the first transcription through images. It is as if the overwhelming power of the dream requires multiple languages.
The dream and the underworld
It is no coincidence that confessional poetry, one of the most important literary exploits of the 20th century, was born precisely from the overflow into the literature of the dream symbols elaborated in analysis.
The very word “nightmare” refers to a precise spatial location: the posture of the evil spirit towering over the sleeper, oppressing them. It is a figure initially explored in images — in my living room, I have one of my favorite 19th-century paintings, A Disturbed and Sleeping Young Woman with a Demon Sitting on Her Chest, a symbol of her nightmare by J.P. Simon.
On my way home one evening, as I passed a statue of Freud, I imagined a contemporary version of a dream journal: an artistic collaboration between me, a dreamer, and an artificial intelligence algorithm, which, receiving my command, would illustrate my dreams.
Instead of the god Asclepius, who was thought to send the ancient Romans images of dreams, I used a more ordinary tool: the Midjourney app.
It was going to be a study of the boundary between image control (not only content but also the light, lens, and aperture) and automatic generation.
So, I created an account on MidJourney and then an Instagram account to share the images created. I named it “The_dream_and_the_underworld,” after Inanna’s underworld journey.
From the familiar to the uncanny
I look at the grid of photos that I update with each awakening and a thousand questions arise in my mind, which perhaps only time will allow me to answer.
Where might one place this experiment in the uncanny valley, the tool that measures the placement of artificial intelligence on a scale from the familiar to the uncanny. How uncanny is MidJourney?
In the dark mirror of the server, figures emerge slowly, like an unconscious that tirelessly grinds and produces: at first they are Rorschach-like blots, then they become realistic faces, four possible versions that can be broken down into four more variations, and so on, in an endless kaleidoscope that stops only when the user is satisfied.
And so the next question is: if the user’s intervention is only narrative, and not factual — that is, it provides the bot with the compositional and conceptual details of the scene, but does not act on the realization except by interrupting the endless compositional merry-go-round — is it art or not?
Most importantly: could we consider it a new interpretation of dreams, where the setting is no longer the Freudian couch but an app?
I think the question is not so much is art or not? It's more a question of is this image mine or the machine’s, or no one’s?
The questions multiply, more numerous than pixels, but for now it is enough for me to step into the looking glass and lose myself in it.
- AI, Translation And The Holy Grail Of "Natural Language" ›
- AI As God? How Artificial Intelligence Could Spark Religious Devotion ›
- Cool New Tool Or Flat-Out Theft? Artists Split Over AI Revolution ›