GAN Physarum: La Derive Numerique

A trained machine learning algorithm that behaves like a living Slime Mould and depicts the future of a bio-digital autonomous Paris.

A GAN (Generative Adversarial Network) is an algorithmic architecture that creates new generative models using deep learning methods. This powerful form of Artificial Intelligence has been trained by ecoLogicStudio’s bio-computational design team to “behave” like a Physarum Polycephalum, a single celled Slime Mould.

GAN-Physarum: la dérive numérique, AI video and biopainting, within the exhibition “Réseaux-Mondes” (Worlds of Networks) at the Centre Pompidou in Paris

The trained GAN-Physarum is then sent on a computational dérive on the streets of Paris. In an AI generated video titled GAN-Physarum: La dérive numérique, it shows us how to decode and reinterpret the patterns of contemporary Paris’ urban fabric.
We witness a transition from the original morphological order to an emergent distributed network of path systems. It is the blue-green, wet and living infrastructure of the city of the next millennium.

biopainting
biopainting

The video is accompanied by the corresponding bio-painting, of 1x0.8 m in size, where a living Physarum Polycephalum stretches its networked body to feed on a grid of nutrients, distributed on the canvas to accurately map the Parisian current
biotic resources. The resulting convoluted network of traces on the canvas is the material embodiment of Physarum distributed intelligence and a nonhuman insight into Paris’ own evolving biotechnological brain.

When the trained GAN_Physarum is deployed it actualises the capacities of Physarum Polyclephalum in solving problems of urban re-metabolisation, carbon neutrality, energetic self-sufficiency and increased biodiversity. Most importantly these ambitious objectives are tackled within a non-human conceptual framework therefore opening up a whole new palette of potential design solutions in the urban realm.

DeepGreen: Urbansphere

DeepGreen: Urbansphere, explains in detail the workflow underpinning the GAN-Physarum algorithm and describes the studio's research into the application of Artificial Intelligence to develop new blue-green masterplans for contemporary cities.

Reinterpretation of the munipal waste collection networks of Guatemala City using the GAN_Physarum algorithm; algorithm training based on Physarum Polycephalum behaviour.

DeepGreen is a long-term project developed by ecoLogicStudio with UNDP (United Nations Development Programme) combining AI and big data analysis. The project aims at designing systemic cities that use their size and collective energy to offer refuge for both humans and displaced wildlife, that promote the emergence of positive microclimate, that replenish depleted water sources and that restore degraded terrains, pushing back on processes such as desertification, land erosion and contamination.

Redefined morphology and materiality of two overlapping systems : local to municipal waste collection networks and the vegetation network in the city of Guatemala City; algorithm training based on Physarum Polycephalum behaviour.
Redefined morphology of local to municipal waste collection networks in the city of Guatemala City ; algorithm training based on Physarum Polycephalum behaviour.

Learning how to interpret large remote sensing data sets from the unique perspective granted by GAN_Physarum, enables a deeper enquiry into the contemporary significance of traditional planning concepts such as zone, boundary, scale, typology and program.

Ultimately with the integrated use of remote sensing, big data analysis and Artificial Intelligence, DeepGreen can be deployed to assess urban vulnerabilities and find specific urban design solutions to achieve immediate and long-lasting impact.

DeepGreen_Satellite view with vegetation in Guatemala City
AI generated view of future blue-green plan of Guatemala City

GAN-Physarum: la dérive numérique and DeepGreen question traditional planning concepts such as zone, boundary, scale, typology and program. Such outdated notions constrain the emergence of a truly systemic approach to urbanisation.

1 / 6

Related press and projects

Project by
ecoLogicStudio (Claudia Pasquero and Marco Poletto)
Project Team GAN-Physarum: la dérive numérique. Biopainting 2022
Claudia Pasquero, Marco Poletto with Greta Ballschuh, Sheng Cao, Alessandra Poletto
Project Team GAN-Physarum: la dérive numérique. AI video
Claudia Pasquero, Marco Poletto with Joy Boulois, Korbinian Enzinger, Oscar Villarreal
Project Team Deep Green: Urbansphere 2021
Claudia Pasquero, Marco Poletto with Thole Althoff, Michael Brewster, Xiaomeng Kong, Stephan Sigl, Eirini Tsomoukou, Lixi Zhu
Academic Partners
Synthetic Landscape Lab at Innsbruck University, Urban Morphogenesis Lab at the Bartlett UCL
Commissioned by
United Nations Development Programme (UNDP)
Photographer
©NAARO