Concept
Objective
The objective of this case study was to explore the possibilities of integrating generative AI in the concept development process for urban developments. After initial contextual inquiry, we found an opportunity for locals and tourists to play a more direct role in contributing to research for architectural firms. On the flip side, architectural firms would have access to live and evolving data to help inform their decision making throughout concept development.
Intersection of Generative AI and Architecture
Omnico is an AI-led architecture firm that uses machine learning to visualize urban futures using a globally crowd-sourced database. Anyone can contribute through Omnico's co-op ownership model. From its continuously growing database, Omnico generates concept designs from curated datasets reflective of the local interactions and relationships between people and buildings within specific environments.
How it Works
User Scenario
The following scenario illustrates the interactions between Omnico's clients, architects, and research contributors. Clients would come to Omnico with an initial request for a build, while architects consult and diagnose needs. Meanwhile, research contributors would gather data points by submitting photos of the physical environment. These data points would help inform material resource planning and architectural drawings.
Design System
Iterations
Low Fidelity Sketches
The scope of this prototype focused on the home page, model gallery page, model page, and individual project page. For the home page, clients must be able to easily access Omnico's best work. For the project page, clients may quickly filter through different models based on their build's needs. When viewing a model, users may view the chosen images within the dataset and access variations of project outputs. After selecting an individual project, users may read more about the build's details.
Mid-Fidelity Wireframes
The main challenge was finding a way to showcase images and display the hierarchy of what it represent. Images fall under the following types: (1) an individual image within a data set or (2) an output variation from a model. I played with image sizes to differentiate between the two types.
High Fidelity Mockups
Figuring out use cases when users are interacting with images and maps was the biggest challenge. An interactive map give users a high-level understanding on where the dataset is derived from. When inspecting a single image from a data set, users would be able to access key details such as the time the image was captured, climate data, physical build attributes, and materials used.
Lessons Learned
Ethics of Generative AI
Through this case study, I learned about the privacy issues that may arise when growing Omnico's database. Such issues include whether or not contributors are able to take photos of certain locations depending on whether it is public or private property. To advance this project, I would look into designing a vetting system that ensures privacy protection and fair assessment of photo quality and ownership value.
