Diffusions in Architecture: Artificial Intelligence and Image Generators
by Matias del Campo, Lev Manovich
The Labors of AI
Sandra Manninger
An in‐Depth interrogation of Labor in the context of dataset creation for Diffusion Models
Have you ever considered how data comes to be? Or, have you ever thought how architecture can avoid the pitfalls that happened historically in the creation of datasets? A crucial component so urgently needed to contribute qualitatively, inclusively, and ethically to current developments in architecture design. In this text, I try to unfold the ontology of data (how does it come to be?), and the epistemology of datasets in this world, where datasets and machine learning (ML) applications increasingly define how traffic flows, goods are delivered, bank loans are decided, stock is traded, and parole is given. Does data appear out of thin air? Does data emerge in a spontaneous generation, perhaps? Aristotle was wrong in his belief that life can spontaneously emerge from inanimate matter22 – the same as data does not emerge out of nowhere. The creation of these datasets involves labor. Human effort. Work. Like every human work, it involves flows of capital, cultural biases, and questions of ethics. For the architecture discipline, current tech companies can be described as anti‐role models. Why? Current artificial intelligence (AI) tech companies rely heavily on surveilled workers such as content moderators, data labelers, shared ride drivers, warehouse stowers, pickers, and packers. Some startups are even hiring people to pretend to be an AI systems (!) like ...