Page 62 - JCAU-6-1

Journal of Chinese
Architecture and Urbanism Exploring abduction in regenerative design

PP is a uniquely peculiar single-celled organism whose
cognitive capabilities have defied scientific scrutiny for
decades, lending it an almost magical aura. Despite lacking
a brain and nervous system, PP exhibits a capacity to
“think” and respond to environmental changes. It builds a
distributed external spatial memory by secreting chemicals
and accumulating traces in its surroundings. In addition, it
possesses an internal temporal sense, allowing it to predict
certain periodic events. Therefore, PP has been dubbed an
“unconventional general-purpose computer” by computer
scientist Andrew Adamatzky (Pasquero & Poletto, 2020).
PP can form networks based on the balance of various
nutrients. In an experiment, researchers strategically placed
food in corresponding locations and prompted PP to find
the most efficient path between multiple food sources.
PP initiates the process by establishing pseudopods in all
directions. Subsequently, branches in areas without food
are gradually abandoned, while those connecting to food
are reinforced and become thicker (Adamatzky, 2010).
This dynamic interaction results in intricate route patterns
and connection structures in PP, emerging from “billions
of dynamic interactions” (Adamatzky, 2019, p. 102).
In “GAN_Physarum: la dérive numérique,” a design
proposal developed within DeepGreen, a satellite image
of Paris undergoes processing to extract the biotic layer
of information. This information is then remapped onto
a physical grid to provide an accurate distribution of
biomass density. Density percentages are subsequently
translated into nutrient quantities on a canvas. The bio-
computational process is initiated with the introduction
of PP (Pasquero & Poletto, 2021a). As the PP grows and
reacts with the substrate, networks gradually emerge,
exhibiting a prototypical path system for the future of
Paris (Figure 4). The bio-computational results of PP
undergo further manipulation through the CycleGAN AI
protocol, visualizing a novel urban fabric across various
Figure 3. Diagram of DeepGreen workflow. Source: ecoLogicStudio, scales, with the resolution incrementally increased. The AI
Deep Green, 2021
eventually bridges the abstract pseudopod networks and
the vivid urban satellite maps, completing the simulation
that might otherwise be overlooked in traditional planning of the future urban morphology for Paris. This simulation
approaches. In addition, in network analysis, the notion transcends “traditional planning concepts such as zone,
of zoning is discarded. The protocols involve a cross- boundary, scale, typology, and program” (Pasquero &
referential analysis of the pathway networks of architecture Poletto, 2020, p. 136). The resulting urban morphology,
and the green system, revealing an intricate interplay shaped by the interaction between architecture and the
between the two (Pasquero et al., 2019). green network, serves as a speculative model for future
These maps subsequently function as training datasets urban planning reference (Figure 5).
for GAN_Physarum, a specific bio-digital algorithmic In the DeepGreen protocols, urban data are sourced
model. At its core, this model is designed to train a GAN from remote sensing satellites, GIS (geographic information
(generative adversarial network), a type of AI, to “behave” system), and DEM (digital elevation models) (Pasquero
like a PP (Physarum polycephalum or slime mold), a form & Poletto, 2017). This information is then processed and
of biological intelligence (Pasquero & Poletto, 2023b). recreated through the coupling of biological and artificial

Volume 6 Issue 1 (2024) 4 https://doi.org/10.36922/jcau.1084