Assistant Professor at Florida Atlantic University School of Architecture, founder of Nonstandardstudio and a PhD student in Vienna, Daniel Bolojan researches AI‐augmented systems that have the potential to liberate architects’ and designers’ creativity. This is not to say that human imagination and machine learning should be equated; they are very different in the way they function and respond to the world, but when combined they offer new creative opportunities.
In the office of Coop Himme(l)blau in Vienna, a series of software machine‐learning protocols and practices continually reinterpret the firm's past and future projects, unleashing new work and possibilities for architectures yet to exist. Here Wolf dPrix, Karolin Schmidbaur, Daniel Bolojan and Efilena Baseta explain the uses and some of the outputs of this suite of tools.
This project examines how interconnected artificial intelligence (AI)-assisted workflows can address the limitations of current language-based models and streamline machine-vision related tasks for architectural design. A precise relationship between text and visual feature representation is problematic and can lead to "ambiguity" in the interpretation of the morphological/tectonic complexity of a building. Textual representation of a design concept only addresses spatial complexity in a reductionist way, since the outcome of the design process is co-dependent on multiple interrelated systems, according to systems theory (Alexander 1968). We propose herewith a process of feature disentanglement (using low level features, i.e., composition) within an interconnected generative adversarial networks (GANs) workflow. The insertion of natural language models within the proposed workflow can help mitigate the semantic distance between different domains and guide the encoding of semantic information throughout a domain transfer process.
"The incorporation of deep learning models into architectural design poses challenges, despite their potential to inform new design processes. One of these issues is the oversimplification of the design problem when employing a discrete, single AI model to tackle a multifaceted, complex design activity. Importantly, the investigation of DL-driven systems requires the identification of components (parts) and relationships between these constituents of this new design workflow. The need to investigate a functional AI-driven design workflow structure with design intentions encoded and incorporated into a human-supervised process is an additional important issue identified by this research. How can specific levels of agency be identified and made explicit in the context of interacting with learning systems? This study investigates a novel human-AI collaborative workflow that combines machine and designer creativity within a comprehensive framework. The focus was on developing a design system, a ""prototype"" with interconnected AI and agent-based models (ABM) to address multiple architectural systems at various design levels (design tasks, design phases) while enacting the designer's varying degrees of agency. Curation of datasets, network types, and connection strategies are the design intentions when working with AI models. In developing a new design workflow, we employ systems theory and the need to deconstruct the design process into its component parts. Design is considered an ""exploration activity"" because it involves the modification and evolution of both the problem goals (design requirements) and the methods and means used to achieve the goals. The investigation centered on prototyping feasible workflows with the following objectives: (1) establishing successful interconnectivities between multiple DL models within the process to manage architectural systems and layers; (2) emphasizing design agency and embedding intentions within each design task within the process. The proposed prototype was applied to three case studies to demonstrate the framework's potential, evaluate its functionality, and assess the outcomes. The experiment described here followed the format of a three-month project (Figure 1). The framework included the use of DL models for (i) design exploration, (ii) generation, revision, and evaluation, and (iii) project development. To examine different types of DL model connections at a global level, the following strategies were identified and implemented: sequential/(unidirectional); parallel and linear; and branching (design problem is broken down into subtasks defined by separate sets of AI models), branching off, or/and merging into a design solution. Designers are the choreographers of how discrete AI models interact with other discrete AI models and human agents. Using this concept as a guide, we can determine the levels of autonomy that the proposed framework driven by interconnected AI can provide. (a) dataset curation: limited agency; (b) types of networks: supervised or unsupervised (StyleGAN, Pix2Pix, CycleGAN); (c) types of connections and combinations of those connections (sequential, parallel, branching)."
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.