The A. Alfred Taubman College of Architecture and Urban Planning is leading the way in using artificial intelligence in architecture.
“As a society, we are already surrounded by AI tools, such as unlocking your phone with facial recognition or asking questions and getting answers from a virtual assistant like Alexa,” said Matias del Campo, associate professor of architecture and director of the Architecture and Artificial Intelligence Laboratory, or AR2IL.
“AI is now going to change architecture. If we do not engage with it, somebody else will do it for us, and then we have no control over the future of our own field.”
Any AI application is only as good as its data, so how can architects contribute to long-term, reliable AI use in the field? By becoming involved in data input early and often, and contributing diverse and inclusive voices to the datasets.
“If you look at some of the early datasets already out there, you immediately notice that they are not done by architects, because they do not understand the qualities of architecture, such as the difference between a bad house and a good house,” del Campo said.
“This is why we need to contribute to the databases now and make sure they reflect architectural excellence, and also have less of the racial or cultural bias that has become evident in the older databases used in other fields.”
He proposes that addressing potential biases in the field is one of the most important roles architects in business and academia will play moving forward.
Databases will remain vital components as AI expands from so-called “expert systems” that process only the input data to inform decisions, to systems that leverage the multitude of data by searching through it for recurring patterns or features and assembling them into unique designs.
“If we are involved (from the beginning), then we can use them to advance the field and make life better for people,” del Campo said.
An AI revolution
U-M architects — in collaboration with colleagues in robotics, computer science and engineering, and data science — have led the charge at Taubman College with the development of AR2IL, an interdisciplinary lab created to improve the application of AI in research fields of architecture, robotics and others.
This interdisciplinary approach of the lab, paired with diverse perspectives, has led to a valuable interchange of ideas within and across research groups.
AR2IL is bringing together experts from across the field and other universities to discuss this emerging paradigm shift in architecture.
U-M faculty and visiting faculty from Yale University, University of Texas, Texas A&M University, Florida Atlantic University and University of California, Berkeley presented their initial findings in November 2022 at the first U-M Neural Architecture Symposium. Colleagues convened again at the recent Data Justice, AI, & Design Colloquium in February 2023.
Still not quite human
Much of the conversation around this emerging field is centered on the unintended consequences of AI, risk-benefit analysis of the new technology, and the amount of human-labor and oversight that goes into maintaining these AI systems (i.e., ChatGPT) and databases.
Most AI-driven design capabilities are generated by humans without accounting for “lived experiences” that robots simply do not have. For example, speakers at the recent Data Justice, AI, & Design Colloquium discussed the frequent lack of foresight behind certain AI products like the germ-killing robots developed by Carnegie Robotics being utilized at Pittsburgh International Airport to eliminate microbes in high-traffic areas.
A panel discussed how these robots still require monitoring from airport janitorial staff to do things like clean up water left behind in restrooms so travelers don’t get injured during their journey.
“The robots essentially need babysitters,” said Sarah Fox, assistant professor of human-computer interaction at Carnegie-Mellon University.
Architectural applications, like DALL-E, are capable of creating AI-generated images with written prompts, but they still require complex algorithms and heavy human oversight in order to generate appropriate architectural building designs or plans for cities.
On a mission to address the lack of lived experiences among robots, research groups at U-M are hard at work on new technologies, such as autonomous robots capable of performing on-site construction tasks at the U-M Robot Garden, while studying how robots interact with their surrounding environment.
SPAN, the award-winning international architecture practice co-founded by del Campo and Sandra Manninger, architect and researcher at the New York Institute of Technology, is undergoing early-stage research and development into the use of AI technology to teach spatial recognition on job sites and the development of other key lived experiences, like engaging with humans and recognizing their facial expressions.