Michele Rinaldi
Italy
Artist Statement
Description
Latent Xylella is a triptych generated by artificial intelligence that reflects on the ecological crisis caused by the Xylella fastidiosa bacterium, which has devastated millions of century-old olive trees in the Salento region of Southern Italy. The work arises from the tension between memory and loss: the images, generated by a GAN algorithm trained on over 10,000 photographs of diseased olive trees, depict fading forms—impressions of a disappearing landscape. These artificially created trees do not exist, yet they evoke what once was, or could have been, constructing an imaginary visual archive of a lost natural heritage. The triptych, in a secular key, echoes the structure of religious polyptychs to celebrate the olive tree as a symbol of identity and resistance in rural Puglia. Inspired by Bosch’s “Garden of Earthly Delights,” the work explores the future of agriculture and the environment in Southern Italy, moving through the living ruins of a wounded landscape. Balancing documentation with speculative vision, Xylella Latente transforms data into imagery, raising radical questions about the relationship between technology, ecology, and memory.
Process
The impulse behind Latent Xylella stemmed from an urgent question: what remains of a landscape when its living form disappears? Faced with the mass death of olive trees in my native region, I didn’t want to simply document the devastation, I wanted to construct a speculative archive using artificial intelligence. I come from a land where the olive tree is not just a plant, but a cultural and social organism. Watching them die, field after field, generated a sense of mourning and urgency in me. Through this work, I sought to give voice to what no longer speaks, using AI not merely as a technical tool but as a medium of ecological imagination. I wanted to explore how memory functions in the age of climate crisis and whether machines can help us retrieve lost forms. Here, artificial intelligence becomes a way to process grief and suggest speculative new roots: latent space becomes a space of memory, where visual traces emerge to continue telling the story of a silenced landscape.
Tools
To create Latent Xylella, I personally developed a Generative Adversarial Network (GAN) algorithm trained on a dataset of over 10,000 photographs of olive trees infected by Xylella, taken during field research in the Salento region. The model was written in PyTorch and optimized following Green AI principles to minimize the carbon footprint of the training process. Specifically, I used mixed precision training (FP16), a technique that reduces memory usage and energy consumption compared to traditional 32-bit (FP32) formats, while maintaining output quality. The training was conducted at the Google Oregon data center, one of the most sustainable in the world, with a Carbon-Free Energy score of 84%. The creative process unfolded in three main phases: dataset collection and curation; algorithm development and training; and image generation and selection. I manually curated the GAN outputs, selecting those that best conveyed the tension between figuration and dissolution. The final images were composed into a triptych inspired by religious polyptychs, giving the work a symbolic and meditative dimension. The result is a biomorphic, hyperrealistic vision, suspended between the natural and the artificial, an aesthetic of possibility that questions the present by envisioning future forms.
Grit Wolany is a freelance art director and AI scout at Zurich University of the Arts (ZHdK). Her artistic focus lies on explorative work with generative algorithms. Grit teaches GenAI at various universities and regularly holds workshops and lectures.
Sometimes, an artwork speaks quietly—and that’s precisely why it stayswith you. Latent Xylella is one of those works. At first glance, strangetrees. At second glance, a digital ecosystem of memory and loss. This is not aplayful experiment with AI. It is a careful act of archiving a natural worldthat may never return. Michele Rinaldi trained a GAN model on more than 10,000photographs of infected olive trees from Salento, a region in southern Italydevastated by the bacterium Xylella fastidiosa. This bacterium is consideredone of the most destructive plant pathogens in the world—incurable andruthless. In Latent Xylella, the olive tree becomes a symbol ofresilience, cultural identity, and the stories rooted in the land itself. Thework is not only conceptually thoughtful, but also technically precise. Insteadof using current diffusion models, Rinaldi deliberately chose a GANarchitecture, trained exclusively on photographs taken during his personalfield research. The process followed the principles of Green AI: smallermodels, energy-efficient processors, and consciously selected data centerspowered by renewable energy. Latent Xylella poses a simple but profoundquestion: what remains when something disappears? And can technology help usmake the invisible visible? In this case, the answer is not a boldstatement—but a quiet melancholy. Perhaps that is what art can offer today:not explanation, but resonance.
