Flore Perdue

Flore Perdue is the 2nd chapter of 'Once Upon A Garden', a project started in 2022 that depicts a world where humans have to live with simulations of flora because it is extinct. The project asks whether the preservation of nature is compatible with the goals of industrialised and industrialising societies, and in general is a reflection on what gets left behind in our pursuit of progress.

On Flore Perdue from Linda Dounia

In the first chapter, a GAN was trained on thousands of images (real and imagined) of extinct and critically endangered flora species from the Sahel region of West Africa. I compiled a list of over 100 flora species of interest from the IUCN's Red List of Endangered Species. For each species, I looked for available images online and within the archive of photographs I took during my travels around Senegal over the years.

To my surprise, for most of the species I was looking into, there were no photographs available anywhere online. So I turned to national botanical archives and was able to find them preserved in herbariums, many dating back to colonial expeditions. I took the herbarium annotations describing the flowers and used those as prompts for DALL.E to imagine what they might have looked like. This is why the GAN for this project was trained on *real* and *imagined* flowers.

Flore Perdue comes two years after the first chapter of ‘Once Upon a Garden’. In this second chapter, I am interested in how much generative AI has grown in its ability to help us remember in more vivid details. I took the outputs of the GAN from the first chapter and used the same annotations I had originally used from herbariums as prompts in Midjourney (V6.0). Flore Perdue features a shortlist of 50 flora species from the initial list. The flowers, much less spectral and more intricate and detailed than the previous chapter, are then combined with a generative algorithm (made in p5.js) whose aim is to obfuscate some of this detail.

This is because while we can feel more kinship with the definition of the flowers from this new chapter as opposed to the lossy flowers from the first chapter, it would be dangerous to lose sight of the fact that neither the detailed nor the lossy flowers actually exist. Both chapters of this project are evidence of an autonomous system, AI, that is getting exponentially better at helping us imagine, but neither can be mistaken for reality.

I refer to ‘speculative archiving’ when speaking about this type of project. It’s archival in that it seeks to preserve what we know about flora we have lost. However, because we are faced with a near complete loss – we have no recorded photographic images of these flowers - we must use speculation to try and fill the gaps with imagination. Through this project, I felt myself over the years negotiating both grief and hope – grief in what we lost and will never actually know, and hope that we’re building technology that can help us remember. I hope that ultimately, the experience of remembering through Flore Perdue helps us feel closer to nature and makes us better, more responsible, stewards of the future of our living systems.