The Nest

Description

This week was a collaboration between myself and Mati to create a data portrait of NYC using publicly available data. We have both been interested in slime mold simulations and their huge potential to create generative form.

Real-life slime mold has been shown to be able to solve many different optimization problems, such as mazes and even the recreation of the Tokyo subway map. Slime mold simulations have been used on everything from art creation to galactic filament simulations to solving the traveling salesman problem.

Mati and I thought it would be fun to use slime mold to interactively solve problems related to NYC's datasets. We chose subway stops as our first test, as it would be a good opportunity to allow participants to re-arrange the subway lines we have all become so accustomed to.

Design Process

We split the problem into two parts; the slime mold simulation, and the map of the subways. Mati worked on creating an interactive map of the subway using his own libraries, while I worked on layering an existing slime mold simulation to have colors and work well with external outputs such as map points. I have built slime mold simulations using compute shaders before and found the p5.js version to be easy to work with. I took the above version and modified it to take in a "food" texture (which in our case is NYC's subway stops); the output does not contain the food texture by default, which it purposeful as we want to see what the slime mold converges to with or without the food texture showing up. Finally, I created a function to generate a gradient texture so that the slime mold can have more colors, and worked with Mati to put all of our work together.

Reflection

The results were interesting, and reminded me of a kaiju eating Tokyo. Still, we can certainly use more features to make this more interesting. We can get better simulation results by relying on the GPU rather than CPU, in this case WebGPU. We can also choose the colors better, perhaps making the map monochrome so as to not conflict with our own color schema. Finally, interaction could be more interesting. For the purposes of an installation, we could rely on Midi controllers that the audience can maniuplate, and can also project the results to be much bigger for when we apply the project to NYC data week.