IFCS REDUX

PROJECT INTRO

Institute for Controlled Speleogenesis (2014 | 2024) by John Becker/WROT Studio and Geoff Manaugh/BLDGBLOG

The Institute for Controlled Speleogenesis (John Becker/WROT Studio and Geoff Manaugh/BLDGBLOG, 2014 | 2024) explores the design and operation of an experimental facility for generating artificial caves in limestone at a variety of spatial scales. The facility relies on an environmentally stable rock-dissolving acid mix, no more toxic than ammonia, applied through an underground network of drip-irrigation lines, storage bladders, and surface injection-arrays. The resulting caverns can then be used for studying and generating novel medicines, including speleo-pharmaceuticals; developing high-density underground agricultural systems; testing recreational caving gear; and prototyping new technologies for subterranean navigation, communication, and mapping. The images seen here—created using a series of Large Vision Models—revisit and expand upon a project first proposed in 2014.

2022 | REDUX V 1.0

Over a decade after the project was started (Institute for Controlled Speleogenesis), I revisited the work with the intent to recreate the concept through newly released AI image generating software. Using a simple set of command prompts meant to guide the AI towards the original vision, I removed myself from the process as ‘image creator’. True to the original intent, machine learning had now re-envisioned the entire project. Standing in new ground I couldn’t help but use every trick I could conjure to guide and edit the process. Attempting to bend the tools to my aesthetic, learning quickly that the core of my design sensibilities can be summed up in a few simple key strokes. The most compelling elements of AI image generators are their lack of precision, and the unexpected forms and details that emerge as a result. I find the resulting imagery quite seductive. The lack of precision in their ability to tell a story gives you a glimpse into what appears to be a vast and complex world without the narrative to stich it together. There is something to take away from the things unsaid in these images. Leaving your audience with enough space to fill in some of the blanks with their own imagination gives the work a breathe of fresh air.

2024 | REDUX V 2.0

After receiving an invitation to an contribute to an exhibition on the use of AI in architecture; my partner for the original project, Geoff Manaugh and I decided to revisit the project for a second time. In the 2 years that had passed an exponential leap in AI tools had taken place. It was now possible to more finely guide the results, make targeted edits to generated images, and achieve far greater accuracy. The resulting images no longer feel like concept sketch, but reach a high level of photorealism.

Obtaining these results comes at a cost. The time invested in creating, refining, and editing these images takes more time and resources, but the end results are undeniably improved. When compared to the original project completed using conventional 3d modeling and 2d digital linework, the time it took to complete the project dropped from roughly 1200 hours spread out over a few years to around 60 hours. The reality is that the original massive effort in man hours was invested largely in thoughtful design work that can not be replaced by AI as I write this. The power of these tools still remains in the ability to produce hundreds or thousands of iterations, culling down to only the best, and using these as highly developed concept images or references for design work. These tools have become a form a sketching. A way to quickly jot down or loosely explore an idea. As these tools continue to evolve the way they impact the design process will unquestionably evolve as well.