YSoA’s Early Lead in AI

YSoA’s Early Lead in AI


As institutions ranging from corporations to world governments increase their investments in artificial intelligence, faculty and staff at YSoA have been ahead of the curve. Led by Vincent Guerrero, senior director of Advanced Technology, and deputy dean Phil Bernstein, developments at the school have included adding AI to the curriculum, investing in new AI tools for students, and collaborating with instructors and administrators across the university.

Shortly following the AI “big bang” in large language models (LLMs) of November 2022, staff in YSoA’s Advanced Technology (AT) department began experimenting with ChatGPT.

One of the team’s early efforts was to address round-the-clock support for students necessitated by charrettes and pre-review production in studio. By final reviews, in April 2023, the Advanced Technology team became one of the first groups at Yale to build an AI tool.

This was to become Minerva, a chatbot named after the famous plaster statue in the fourth-floor pit, which students can use to answer common and emergent questions about advanced technology operations such as, “Can you cut mirrored materials on the laser cutter?” Minerva boosts student production capabilities while freeing up the AT team to pursue the development of more student resources and innovation, especially during peak studio times. Minerva is intended to be consulted before students turn to creating a support ticket for their issues. If necessary, students can escalate their questions with the command, “I want a human.” Resolutions to issues that Minerva fails to address are then added back into the chatbot training database.

Soon after, Phil Bernstein challenged the team to build something even more ambitious: a tool that could be used for research projects in the form of a custom chatbot. Called Vulcan, it was meant to be trained solely on project-specific data. For example, Vulcan could be given context on and information about accessibility codes and then be asked to assess accessibility compliance for a given scenario. YSoA’s Vulcan came online well before OpenAI’s custom GPTs with similar capabilities.

Bernstein had already been pursuing his own research into AI and its potential ramifications for the practice of architecture. His book Machine Learning: Architecture in the Age of Artificial Intelligence (RIBA, 2022) outlined the opportunities and potential threats of this technology for the architecture profession. In Spring 2023 Bernstein delivered an updated talk together with Sam Omans on the subject as part of the School of Architecture’s event series “Faculty Work.”

The university has also begun developing custom AI tools to meet demand across various schools and units. The new Clarity platform, a general-use chatbot created specifically to meet requirements for handling sensitive data, has been launched university-wide. Vincent Guerrero has been active on the Clarity program advisory board and as a project partner. Part of his role is to integrate developments in AI with emerging needs in the school curriculum. He also worked on the team that launched the university’s AI website, intended as a hub for guidance and tools including updates, training, interdisciplinary collaboration, and AI-related events.

Members of the YSoA faculty have been adding AI to the curriculum through seminar courses. In Spring 2024 Brennan Buck introduced the use of Stable Diffusion (free and open source), via his course “The Black Box: Architecture in an Age of Opacity.” Students were tasked with diagramming their design processes and then documenting the influence of AI on them. Buck along with Bernstein and Omans’ supported by Guerrero, are offering “Scales of Intelligence: AI, Agency, and Architecture” this semester. The course starts with the recognition that “unlike predecessor tools like CAD or BIM, this new class of tools acts with independence, challenging the roles, responsibilities, processes, and fundamental agency of architects and the systems in which they operate.” Students will develop “an understanding of the trajectories of AI technology, interrogate the process implications of autonomous generative algorithms, and explore larger socioeconomic and ethical questions as machines fill larger roles in the discipline and the society it serves.” Will AI render architects more or less relevant? Bernstein and Guerrero have both hosted AI info sessions on similar questions for interested YSoA alumni.

What is the future for AI at YSoA? Guerrero and his team are working on tools for converting sketches into images and images into other images using Stable Diffusion, an area of investigation where YSoA maintains a lead over the rest of the university. Sustainability is also a big concern overall. New graphics cards in YSoA computers allow for local computing, potentially reducing the amount of electricity needed for AI processing.

Justin1
Justin Levelle, Pre- and Post-AI Score, final project for The Black Box course, taught by Brennan Buck.