Diffusions in Architecture: Artificial Intelligence and Image Generators
by Matias del Campo, Lev Manovich
Five Points of Architecture and AI
Andrew Kudless
In 1923, Le Corbusier proposed his Five Points of Architecture, a set of formal strategies rooted in the socio‐technological changes of that era. One hundred years later, we face our own social and technological challenges, ranging from the climate crisis and inequality to the emergence of artificial intelligence (AI) and automation. Unlike the International Style that arose from Le Corbusier’s Five Points, today’s implications are less formal and more disciplinary. The profession and academia are faced with the need for transformative changes that redefine the designer’s role in contemporary society. Specifically, the integration of AI demands that designers become more critical of media and its collective creation, manipulation, and dissemination. The following new Five Points summarize the opportunities and challenges ahead as we move toward an understanding of architecture in solidarity with both human and nonhuman beings.
The Challenge of Bias
Large Language Models, including Latent Diffusion Models that rely on training data scrapped from the internet, are as truthful as the underlying data. These models are essentially statistical predictors: based on the patterns learned from the training data, the model responds to a prompt. Like us, the models are products of their training and are fallible, subjective, and incomplete in their knowledge. All types of biases, from the benign to the insidious, are inherent in systems ...