Diffusions in Architecture: Artificial Intelligence and Image Generators
by Matias del Campo, Lev Manovich
Ontology of Diffusion Models: Tools, Language and Architecture Design
Matias del Campo
Diffusion models69 are a family of deep generative models that have surfaced in recent years. They can be considered the state‐of‐the‐art deep generative models and have surpassed the abilities of the previously dominant generative adversarial network (GAN),70 in particular regarding the demanding task of image synthesis.71 Apart from image synthesis, diffusion models can be applied to tasks such as machine vision,72 natural language processing (NLP),73 robust machine learning, and temporal data modeling.74 Apart from pure applications in computer science (CS), there are multiple projects operating in an interdisciplinary fashion to achieve things like medical image reconstruction,76 computational chemistry,77 and architecture design.78 This section provides an abbreviated overview of the currently existing diffusion models, their mode of operations, and categorizing the different approaches. This overview serves as the launching pad for a thorough discursive interrogation of ontological and epistemological questions emerging from this novel tool of design.
When talking about tools, we will dive into some concepts regarding the cultural consideration of tools, contrasting with the purely technological nature of the introduction, but expanding it towards the meaning of tool in the context of learning systems. Videlicet, we will be discussing tools that learn. Text‐to‐image models refer in ...