On the dialectical nature of dreams

A conversation with aurèce vettier.

“He walked against the shreds of fire. They did not bite his flesh, they caressed him and flooded him without heat and without combustion. With relief, with humiliation, with terror, he understood that he too was an appearance, that another was dreaming him.” [1]

It is exactly 9:23 a.m. on Monday, October 3rd. We are at the studio of aurèce vettier, in Ivry-sur-Seine, a sacred space, quite neat for an artist's studio I must say, with the perfect dimensions, with the perfect temperature. There is something “auratique” about this place. There is a quietness in the air, but at the same time it is a little sad. The fog that floods the city of Paris on Monday morning does not let us see the Tours Duo that rises behind the studio window. Maybe it is better this way, maybe the industrial perspective has nothing to do with the romanticism of a space like this. The fog settles better to the eye.

Paul prepares a coffee—decaf. I stand in front of a group of 33 small paintings, unrecognizable to me, all new. I don't understand much. A fish cutting off its own head, a cat—or at least a cat-like shape—on a roof, a bearded man, several times, the same bearded man. Again, the cat, a black cat, bucolic, a forest, several forests, a rising hand holding a strange creature. A fish? Again, the fish... I don't understand anything, but I start to cry. I start to cry because those paintings remind me of the loss of a friend I had exactly 3 days ago, far away, in Cuba, a friend I could never say goodbye to. Indeed, now that I remember I haven’t stopped crying for the last three days. Those same paintings that I stare at fixedly, without moving, while I cry, make me regret how insignificant we are, how tiny. They make me think of the things we don't know, of what is not given spontaneously, of what we will die without knowing it never existed. Those paintings, at that moment, are a penetrating knife.

I sit at the small table in the center of the studio. We have another coffee—decaf again—but I need to understand what is going on, so we start the conversation.

When did you start having these dreams?

The arrival of these dreams was related to moments of transformation and alteration in my personal life. It started about two years ago. I started to deal with new things for me, new and unexpected ones like anxiety, to have prolonged moments of anxiety. In addition to that, after I turned 30, I started to frequent the so-called existential crises, mainly associated with the passing of time and the dizzying way we live.

However, I began to realise that what was happening to me was not specific to me. I realised that others had already experienced this torment, and seemed to have broken free from it, which meant that there was probably some sort of recipe for moving on, although I understood that the process was long and potentially complex.

When you went through these moments of anxiety or distress, didn't you feel as if they were eternal, as if they would never end?

Absolutely. But at the same time I began to feel more and more that I had to find a way to deal with it. I started working with therapists, and then with a magnetizer. I knew an artist who shared some of the experiences I was having at the time and it was she who introduced me to the magnetist. His job is to unblock energies. For me, this type of exercise represented an attempt to establish a new structure in my life, more solid on a personal or emotional level, and therefore on a professional level. In the first few weeks of my work with the magnetizer, I began to have different types of dreams. Each time they were very photographic dreams. They were not, for example, erotic dreams or nightmares. The whole thing was sometimes blurred, but I could almost always describe or remember at least one or two scenes from these dreams.

I remember that at that time I was working on Potential Herbariums, and I also had many ideas about what I wanted to do next, but I suddenly felt the need to pay more attention to the dreams that were occurring in my daily life and were becoming very recurrent. Furthermore, as I had already been working for months on artificial intelligence (AI) algorithms capable of transforming text into images, I thought it would be a good idea to use this algorithm to create extremely personal images. Given the difficulty of representing them as they are because of their blurred or ephemeral nature, I could not develop such a representation myself. So I decided to use AI as a way of working and as a way of pushing my creative boundaries.

Since we are talking here about the representation of dreams, and the different ways in which this representation can be deployed, or even concealed, we could argue, at first glance, that this series of dreams evokes certain aesthetics of art history, such as surrealism. However, in the case of Circular Ruins, there is something beyond surrealism—both aesthetically and conceptually. I see this series of paintings more as an arsenal of images that have more to do with the future than with the past, while at the same time being very much part of art history. In almost all your work, it's about giving an aesthetic and artistic sensibility to a certain amount of information—a dataset. Is this the case here?

Everything can become information: a song, an image, a certain iconography, everything is ultimately data. In this sense, I am aware of what movements like surrealism have meant for the history of art, but I don't see this series as a specifically surrealist set of paintings. Rather, for me it is a continuation of the generative art branch, an area I have been exploring in my work for some time, interpreted in this case on a more personal level, as it is about dreams. In fact, I have often wondered whether this series should be shown to any audience at all. And if so, I have often wondered how to present it to that audience. Did I necessarily have to explain each of the paintings? Should the images be of a higher or lower resolution?

The general aesthetics of the paintings are affected by specific constraints related to their production. First, the tool with which they were made is an algorithm that I assembled and tweaked with personal elements in 2021. I adapted it to my aesthetics, since it was important that this algorithm fit in with the rest of the visual ecosystem I work with. Moreover, the algorithm had internal limitations and was not yielding high resolution images at the time. However, it was enough to work and create images with fascinating main scenes and backgrounds.

The day I finished producing this series, I think it was the first time in my short career that I felt I had nothing to add, conceptually speaking. I felt like everything had been said and done. I had the feeling that I was permanently extracting some of my sadness from my body. I didn't want to exhibit these paintings immediately after I finished them for a simple reason: I wanted to live with them, work with them, breathe with them. This series is a kind of visual album of a stage in my life that was full of recurring questions and anxieties. It therefore represents, if not everything I am, at least a set of remnants of striking experiences during this path of the cross.

So, we could say that this series is also an expression of fear, seen from different facets.

Absolutely, but I hope it is not seen as a paralyzing fear. For me fear is something to go through, a taboo behind which an oasis can be found. That's how I see it in my life, and I think that's how it can also be seen in this series. Hence, not all the dream scenes appeal to murky or dark moments, but there are other scenes of more clarity and calm such as A French castle and its orchard.

It makes me think of a phrase from Baltrušaitis that goes something like this: "The mirror is the attribute of prudence and embodies wisdom. Reflections in thought and in the mirror are designated by the same word." [2] Seen this way, I would like to think that this series then functions as a reflecting mirror of your own life, which, like all life after all, has its dark sides and others not so much. Is that so?

I think so. There are many hidden elements in this series that function as subterfuges of my reality, that are almost related to a specific moment in my past. For example, very few people know about my passion for rap. In fact, one of the dreams I had and therefore one of the paintings is about French rapper Kalash Criminel. I dreamt that Kalash was trapped inside a Rothko painting, plain and simple. And so, he is represented by the algorithm.

And why the small format? What comfort do you find in it?

Perhaps because I wanted to dialogue, perhaps unconsciously, with the concept of the icon and what it represents. The work of Robbie Barrat, who is a very good friend and a great AI pioneer, is currently focused on the reclamation of the icon in the use of artificial intelligence. If we think about it, this idea is extremely interesting. What is an icon in the age of technical reproducibility, to put it in Walter Benjamin's terms? How can we agree on what is "iconic" and what is not, in the age of over-communication, where so many visual icons try to predominate simultaneously? This question alone could be the subject of another interview.

Anyway, the small format also answers a simple need: practicality. I am also an art collector, so I have collector's concerns in my work: for what purpose the work should be acquired, where it could be placed, etc. The paintings in Circular Ruins don't need to be installed in spectacular spaces, where you gain recognition from your peers. They have a life of their own, they exist on their own beyond social interaction and evidence. They could be installed in a corridor or a more intimate room as their size lends itself to this.

Finally, as we mentioned, the algorithm did not provide very high resolution images, which also helped to limit the size of the final paintings.

How one lives with an art piece as a collector is a crucial question for me, and that is why I am constantly experimenting with new ways of showing the works. For example, for this exhibition I collaborated with bem.builders to install some of the paintings in their metaverse foundation, in a virtual world called Voxels.

It's hard for me to think about this dream series without thinking about how poetry—as a visual element more than anything else—has always been the backbone of your work. Is there a particular relationship between this series of dream paintings and your poems?

In my work there is always a relationship with poetry, although sometimes it is more evident than others. This series of paintings about dreams is no exception. Knowingly it also comprises references to other artists and works of art. One of them is a tribute to August Strindberg for example, while another is the result of a dream I had about Damien Hirst's exhibition Treasures from the Wreck of the Incredible [3]. This Venice Biennale exhibition changed my usual way of looking at art. At first, as I was exploring the ground floor of the Palazzo Grassi, I sincerely and maybe candidly thought Hirst had become an archaeologist. Then I understood, after a while looking at the works, that it was all an invention, a cynical game with the truth. At another moment I dreamt only of David Lynch's name—or at least that is what my memory managed to memorize. And only David Lynch's name was what I offered to the algorithm. And so on.

I would like to know more about the process behind the series…

The meta-process is always the same. My starting point is a personal raw material. It can be a herbarium, texts, any kind of information. Then I collaborate with a machine in the sense that I submit these initial thoughts to the algorithm, either an AI or an algorithmic tool, to push them towards more interesting entrenchments, which coexist between real space and digital or metaphysical space. This is a summary of my general approach.

Now, in this set of works I used a new kind of AI. So far, generative AI could derive images from a data set, after a long training process. Starting from a body of work, the goal of this GANs (Generative Adversarial Networks) is to generate an image that synthesizes a specific idea, just as it was done for the series of Potential Herbariums. While working with this series I started to implement new algorithms. I connected a certain type of GAN (called VQGAN) with another algorithm called CLIP [4], allowing the programmer to type the text—also called “prompt”—and have an image that attempts at representing the text already given to the algorithm.

For the first work I did with this process, I didn't even implement the full algorithm: I did some tests online instead, not on my own computer. I thought: I'll try to represent myself, and I'll try to work with these first images. It's funny how, when you try to describe yourself, to synthesize yourself with a simple group of words, you very quickly plunge into an abyss of reflection. Of course, this was to be expected. But who am I? In my case, I thought: "I am a bearded man working on a completely beige painting". So that was the first image I received from the AI. An image of myself or how I saw myself. I was very moved by this first image. Being an artist is not a function, nor even really a profession, it is rather a vocation that allows you to bring about exactly this type of emotion.

Let me interrupt you there. It remains an enigma to me to understand how, departing from words, it is possible to arrive at an image.

This is also a mystery to me. It is impossible to know exactly for mathematical reasons. That's the magic of it. But I can explain how to set up an algorithm in the first place. How to train it.

First, we work on the text, with enrichment algorithms such as GPT-3 [5]. This allows us to explore different keywords from the basic sentence given to it. Then the second part of the algorithm—the GAN—can generate images based on the information you give it. So, for example, when you ask for a man with a beard, it can find all sorts of images, words and symbols in a dataset that refer to the idea of a man with a beard. Pixel by pixel, the image is constructed, assembled until it optimally matches what the programmer has asked the algorithm to do. So I can't explain how images are formed; they have too many dimensions that are elusive to the human brain.

The advent of artificial intelligence in art is a genuine revolution. It is no longer a question of what machines can do for artists. This question was already explored in the 1960s, when artists such as Herbert Franke, Manfred Mohr or Vera Molnar questioned the relationship between the artist and the machine. At that time, artists demonstrated that the machine can expand formal possibilities.

Today, anyone, whether they come from the art world or not, can use AI and generate images very easily. So the question changes drastically: it is now a question of knowing what the machine cannot do for the artist and, beyond that, for the human being in general. From my point of view, what the machine cannot do is create unexpected bridges between completely different universes and inspirations, or select interesting media such as bronze or painting—to give just two examples.

In short, the machine already knows everything, it has all the references in it. We have to go and dig them out. This is the artist's job, more than ever.

It is only in the process of selecting the media, the concept, the story and the visual ecosystem that the difference can be found. It is the artist, a.k.a. the programmer, who through his or her culture and disruptive ideas, can unfold something innovative, continue the research. For me, it was particularly revealing to work for these paintings with a not so sophisticated algorithm. It allowed me to find more possibilities to be creative with the tools I had, without having to be so predictable with the machine, just like when you take pictures with an old camera with limited functions. I was also able to continue to explore the question of sustainability: how do you create a physical and sustainable object from an eminently virtual process, based on digital data?

This data cannot be touched, a priori. This is why, in my work, I insist so much on making art that is palpable, not constrained by the screen or the cloud.

What I also like about the field of art creation in relation to artificial intelligence is the possibility to contribute to building a community, through platforms like Twitter or Discord. The artists currently working with artificial intelligence all know each other, participate in exhibitions that I think will one day be crucial, and we collect each other.

Of course, the generative artists of the time did the same in their own time—I am thinking, for example, of François Morellet's extensive links with his American and other international counterparts, which are clearly visible in his personal collection. In our field, artists are not afraid to exchange their "recipes", to exchange their knowledge: this is the Internet culture.

I think that since the invention of photography, this is the first time we have experienced such a paradigm shift in art. And I'm not talking about NFTs at all, these are just a channel, a way to "play with the community". No, this change is accelerated by AI, because it is a kind of intelligent brush that expands the fields of knowledge built up to now by humans, and that allows it to go beyond what humans can understand. Thanks to this hybrid between personal concerns and computing power, working with AI opens up new worlds. It also gives more value to the work of some people who are more curators than creators. You don't necessarily have to be a creator to work with this kind of tool, especially when you have a good conceptual vision. I also feel that AI as an artistic medium is beginning to enjoy genuine recognition. I am sure that in a few years artists working with tools that are not yet widespread such as AI will enter galleries and the institutional system in general. Working with AI will cease to be something exceptional. I am already thinking about what will happen after that.

How could we refer to “authorship”—quoting Walter Benjamin once again—when it comes to algorithms? How can you appropriate a certain technology and make it "yours"?

That's a good question. When you create a technology in general what you do is to connect "microservices" together like a patchwork. Each microservice achieves a certain task. And then it is also important—as I mentioned above—the way you train the algorithm. Think that when you start training an algorithm you are in front of a baby's brain, metaphorically speaking. It's a brain with thinking capabilities, not biased or perverted by anything in the Rousseau sense, but you must teach it everything. You must show it the things you did in the past and what interests you in the present if you want to imprint your particular style on it. And if you do it yourself, it's quite an expensive job. It's expensive when you're creating an AI from scratch—that is, when you want to train an algorithm exclusively for yourself, with your own aesthetic. But there are fewer and fewer commercial barriers if you want to run simple tests: there are platforms that allow you to run these algorithms almost for free. The resolution will just be much lower.

Do you expect any specific reaction from the public receiving this series of paintings? After all, you are sharing "your dreams"...

The moment they will be shown to an audience, these works will cease to be mine, these dreams will cease to be only my dreams. I never expected anything from anybody, and this time is no exception.

--

[1] Borges, Jorge Luis. Les ruines circulaires. 1940.

[2] Baltrušaitis, J., Ensayo sobre una leyenda científica. El espejo. Revelaciones, ciencia-ficción y falacias, Madrid, Miraguano-Polifemo, 1978, p. 281.

[3] Exhibition by artist Damien Hirst held as part of the Venice Biennale from April 9 to December 3, 2017.

[4] CLIP is an artificial intelligence model under the form of a neural network. It is trained on 400 million (image, text) pairs. An (image, text) pair might be a picture and its caption. CLIP acts as a bridge between computer vision (images) and natural language processing (text). Among possible use cases, you can input an image into the CLIP model, and it will return for you the likeliest caption or summary of that image.

[5] Generative Pre-trained Transformer 3, known by its acronym GPT-3, is an autoregressive language model that uses deep learning to produce texts that simulate human writing.