Neural Monumentality

Ricardo Guisse + Spencer Reay + Tam Nguyen

In recent years, artificial intelligence (A.I.) has been increasingly disrupting creative fields such as art, music, writing, and graphic design – and architecture is not an exception. As this novel technology makes its way into the design process, our thesis explores how this is leading to the advent of posthuman architecture. Our thesis involves the understanding of the generative methods behind deep learning diffusion models, Midjourney and Dall-E, and their application in the design of a tangible architectural object.

The potential of Midjourney and Dall-E is in their capability of crafting compelling architectural imagery. But as we started to generate these images, we observed that there were some inherent biases towards Western culture. We questioned these cultural implications in the design process, and to explore this hypothesis we decided to look exclusively into Midjourney’s dataset. Our observations revealed the complex composition of the annotated data. Because the internet heavily biases Western content, every step of processing and filtering introduces bias into this dataset, as algorithms and programmers determine the inputs and machine learning continues to excel. We considered this as the biggest caveat of using A.I. in architecture.

To further explore the complexity of Midjourney’s bias, we chose to design a monument; which we’ve defined as an archaic object capable of representing a multitude of different architectural styles and cultures. Our design process involved generating images in Midjourney by prompting basic architectural traits from a variety of cultures, interpreting the results, modeling these interpretations using 3D software, and feeding model snapshots back to Midjourney and repeating the process.

With each diffusion, the architectural and cultural features of the monument began to become estranged and defamiliarized, provoking ideas not previously conceived, challenging our preconceived interpretations of architecture, and ultimately conveying its post-human design aesthetic.

We want to examine the potential that diffusion models have as tools for design, because of their capability of crafting architectural imagery that can be compelling and inspiring.
Phil Bernstein once said that architects and designers have been capable of using automated tools such as CAD and 3D modeling to produce advanced architecture, now the question is:
How can we adapt these novel autonomous tools to do the same? To this end, we theorize that we are at the advent of a posthuman architectural design process.

The focus of our research involves diffusion models, a type of machine learning within the realm of artificial intelligence (AI), and its use in architecture, which is subject of a larger critical interrogation within architecture theory regarding posthuman design methods. While our research may involve terminology and aspects of computer science, our focus only involves its application as a design tool. Other disciplines such as sociology, archeology, anthropology, and world history may be touched upon but only for reference purposes in providing definitions and contextual analysis relevant to architecture.

For more information about this project, please visit https://cdn.me-qr.com/pdf/14326790.pdf