© Pascal Sgro
Introduction by Michel Poivert Who’s afraid of artificial intelligence? Photographers, perhaps—but not all of them! When it comes to AI, photography remains a model. The photorealistic quality of AI-generated images serves as a reminder that visual data production is fed by photographs stored in digital memories. Even if they no longer take pictures themselves, photographers can still create using the entities held in reserve. Beyond the folklore of hyperfakes (deepfakes), which AI-generated imagery is often reduced to, photographers explore the technology’s potential to imagine the world. Dicible/Visible Sayable/Visible The generative AI systems at the heart of the exhibition work on a text-to-image basis. Their power lies in their ability to learn through deep neural networks. Deep learning...
Introduction by Michel Poivert Who’s afraid of artificial intelligence? Photographers, perhaps—but not all of them! When it comes to AI, photography remains a model. The photorealistic quality of AI-generated images serves as a reminder that visual data production is fed by photographs stored in digital memories. Even if they no longer take pictures themselves, photographers can still create using the entities held in reserve. Beyond the folklore of hyperfakes (deepfakes), which AI-generated imagery is often reduced to, photographers explore the technology’s potential to imagine the world. Dicible/Visible Sayable/Visible The generative AI systems at the heart of the exhibition work on a text-to-image basis. Their power lies in their ability to learn through deep neural networks. Deep learning is achieved via Generative Adversarial Networks (GANs), which are machine-learning algorithms. After being trained in data centers, they can generate content from natural language queries (rather than code), known as prompts. But where do these AIs retrieve the billions of data points used to generate images? In latent spaces. The term evokes the «latent image» in analog photography—the invisible image on film that becomes visible in the lab. Latent spaces are invisible realms where each image is associated with a word (indexing). These pairings are encoded and can be located in a multidimensional space through their coordinates. A latent space is thus a dynamic reservoir of data, connectable based on the queries we pose. By principle, only what has been recorded exists in these spaces. However, the combinations are countless and depend on the moment. AI-generated images are akin to snapshots taken within the latent space. Filters impose limits, excluding “inappropriate” subjects for moral or political reasons. Photographers become primarily creators of text (prompts). The language is «performative»: what is said becomes an image. This transition moves us from photography (writing with light) to promptography. The art of description takes precedence over visual creation, reversing the previous order where captions came after visual production. The weight of words makes the sayable the condition for the visible. For photographers, a game emerges in the dialogue with AI, refining prompts to guide the system toward convincing results. Surprises abound, as AI produces numerous absurdities referred to as «hallucinations» or «confabulations.» Much like in psychology, these are memory errors presented as truths, which artists can exploit aesthetically. This is how algorithm-generated images are recognized: they are machines for creating chimeras. Recursivity/Uchronia Generative algorithms allow us to explore our imagination. By leveraging data from latent spaces, they reveal the content of our visual culture. Is this not, in a sense, akin to photographing within our collective memory? Images of images, AI-generated «photographs» mirror our imaginaries. Photographers can thus revisit history and construct new scenarios. A visual score is written from recursivity—a constant return to what is «already there.» How far is this from Roland Barthes’s definition of photography in Camera Lucida—the famous that-has-been? Except that in latent spaces, the experience of the real is already mediated by images and words. Amid a truth crisis, the false and the true have become interchangeable. Photographers choose a third path: that of anticipatory imagination. One might call it «alternative fiction,» where the illusion of truth is not the goal. Instead, it’s about imagining possible worlds rather than escaping reality. Are photographers still photographers, or have they become anticipatory novelists? Their art turns prospective. They adopt the principles of counterfactual history: a logical narrative can arise from a hypothesis. What could have, or might yet, occur becomes the premise for rewriting the world. This often reveals hidden truths and highlights phenomena made invisible by official history. Photography thus uncovers another facet of its power: by venturing beyond the real, it repopulates our imagination and (re)generates the images and narratives we need. AI operates in a temporal mode of uchronia—a rewriting of history through the modification of the past. It is no surprise that some photographers revisit the very history of photography and, more broadly, the concept of the archive. Digital memories, with their infinite labyrinths, harbor potential for phantom-like images. These are spectral apparitions that challenge our relationship with time. The retrofuturistic aesthetic of AI-generated photographs stems from this inversion of temporal polarities. Yet, the plasticity of their forms introduces a novel, “liquid” style, shaped by the data blending intrinsic to digital engineering. AI thus enables a morphogenesis of images, endowing them with lifelike appearances. Utopia/Dystopia? What worlds does AI allow us to imagine? An alienating universe of reconfigured memories, or the emancipatory promise of new narratives? Chimerical beings may be unsettling cyborgs or unruly robots, but they can also reveal fantastical figures that heal our traumas. Photographers play a central role here. Their creations reveal both AI’s potential and its limits, exposing its spectral nature. Those who long served as witnesses to reality through documentary photography have become shamans, guiding journeys through space and time. AI “photography” does not align with the notion of progress. The visual potential explored by photographers does not unveil a future where the world is more modern. Instead, it challenges the very idea of the future. Each artist’s possible worlds are forms of probabilities. What does AI teach us about photography? Perhaps that it was never the art of reality we believed it to be. Post-photographers, neo-photographers, or more precisely, promptographers, these artists offer us an alternative approach to images in the age of the attention economy. For them, looking can no longer occur without imagining. - Michel Poivert, photography historian and co-curator of the AImagine exhibitionOpening Hours
| Monday | Closed |
| Tuesday | Closed |
| Wednesday | 12:00 - 18:00 |
| Thursday | 12:00 - 18:00 |
| Friday | 12:00 - 18:00 |
| Saturday | 12:00 - 18:00 |
| Sunday | 12:00 - 18:00 |