“Self-driving cars can identify objects as they drive,” a video from the company Smartvid.io proclaims. “What if we could bring this ability to the industrial world?” The Cambridge, Mass.–based outfit has developed technology to do just that: It offers software that analyzes huge amounts of data—in the form of photos and videos from construction sites—to identify safety risks that might not be evident to a human observer. It tags, for example, workers who are missing hard hats and types of ladders considered risky, promising to help “reinforce safety culture.”

“The risks might not be obvious right away, but when you look at the total data, it emerges,” says Imdat As, an expert in the rise of artificial intelligence in the field of architecture and founder of Arcbazar, a competition platform for architectural design projects. As notes that this type of artificial intelligence used by Smartvid.io—called deep learning—is an early application of what we’ll see from AI in architecture more broadly, such as computer tools that will offer alternative design solutions.

Many architects are excited about these opportunities, and some large firms are exploring the latest technology. But what about smaller firms? According to the AIA's 2018 Firm Survey Report, 75.8 percent of firms have one to nine employees. How will these smaller outfits, with smaller budgets, confront the rise of AI? Though smaller firms may face resource challenges, as artificial intelligence tools become more widespread and less expensive, they perhaps stand to benefit the most.

From Automation to Artificial Intelligence

Already, architects are increasingly using technology to automate the quantifiable aspects of architecture, such as apps that give a designer almost instant access to zoning rules or building codes in a certain area. But this isn’t AI, explains As, noting that the way we think about AI today stems from work that began accelerating in 2011 because of better and cheaper computers, as well as increasing amounts of available data. “Ninety percent of all data available in the world has been produced in the last two years,” he says.

Artificial intelligence thus doesn’t merely automate a task by serving as an efficient clearinghouse of data; rather, it analyzes data and generates new ideas or solutions, similar to how a human mind would approach a problem. Hence, there is a need for more and better data from which machines can learn.

While most of the currently popular AI applications involve the processing of text, audio, and images—such as what self-driving cars and Smartvid.io’s construction software does—As says new forms of AI tools that can learn from different data sources, such as drawings, are on their way for architects. (Other forms of AI research that are not datadriven, such as evolutionary algorithms, also might someday provide alternative solutions to architectural issues.)

In the future, for instance, architects will likely be able to tell a program that they want a house for a family with two children and a dog that must also be handicapped-accessible. Though the system can theoretically generate millions of examples, it will narrow them down to the dozens that it “thinks” are best, and the designer can further develop one or more of those.

As says that in the long term, these systems might be further developed into consumer products that can automate design tailored to the taste of clients directly. And Ron Beqiri, an architect and spatial planner with expertise in technology who hails from Prishtina, Kosovo, speaks of the possibility of industrial-size, autonomous 3D printers that could then build structures without the need for anyone to manage them—technology he says is currently being studied in the Mars Science City in Dubai, which simulates building on the Red Planet.

What Will AI Mean For Architects?

It’s unclear when exactly such tools will become available. Because deep learning systems demand data that is machine-readable, text, audio, and images lend themselves more easily to current AI applications than do graphics that represent three-dimensional architectural spaces. But scientists like As are at work creating alternative representative models of architecture—such as via graphs— that can fill this role. So while it may be years, or even decades, before machines can design buildings—especially good ones—the technology will sooner or later be upon us.

Though there is much anticipation about such change, many people also fear it, particularly due to its potential for replacing human labor and eliminating jobs. Moreover, while early 20th-century theorists such as economist John Maynard Keyes predicted that by the end of the last century automation would allow people to toil only 15 hours per week, such a vision never came to pass. Instead, in spite of automation, we’re working even more—following an increase in corporate paper-pusher jobs—in jobs that anthropologist David Graeber has termed “bullshit jobs.” Hence, some suspect that AI will simply prompt another round of job loss and replacement with perhaps even more dystopian results of poverty and inequality.

But there are many architects who are more optimistic. After all, architecture—a creative endeavor that actually makes things—doesn’t qualify as a “bullshit job” and is one profession that will thus be harder to automate. As a result, the design tools that As describes might be more helpful than harmful to the profession in the short to medium terms. (Lower-level work at architecture firms, such as jobs that involve billing or other administrative tasks, will likely be more vulnerable.)

“The computer and the program will never be a replacement of good judgment, extraordinary design, and creativity,” says R. Denise Everson, an architect with the firm Cure Architects outside Washington, D.C. “Technology should not think for architects; architects should always be the thought leaders.”

Small Firms: Opportunities and Challenges

Everson’s firm currently employs two architects. As says that automation and AI can ultimately make a firm the size of Cure more competitive because it will have access to the same technologies as larger outfits. Natasha Luthra, AIA, director of the innovation program at Jacobs and the 2018 chair of AIA’s Technology in Architectural Practice Knowledge Community (TAP), concurs. “Small firms have real advantages in certain ways because technology can democratize,” she says, “and small firms can be nimble in their adoption of it, such as by renting software on a monthly basis to see if it works for them before looking to make a big investment.”

At the same time, such democratization won’t occur overnight, and the technology will have to become cheaper. Phillip Bernstein, associate dean and senior lecturer at the Yale School of Architecture and former vice president at Autodesk, says that automation and AI—like every other type of technology we’ve seen—will have to become more established before it becomes available to the wider market. “Software companies or large firms will use the technology first, and it will then filter down to small firms, whose resources won’t allow them to use it until it moves into the middle market,” he says.

Another challenge for small firms: data. Good quality and a large amount of data, As noted, is key for deep learning systems to work well, and small firms may simply not have access to vast quantities of data. As hopes that firms and architects will move toward sharing their data so that the profession as a whole will benefit. “There are some shared project platforms out there already,” he says, “but they’re not big enough or formatted the right way.”

Everson says that she and her partner are already meeting with data scientists who focus on automation and AI so they can determine how to fit them into their practice. “We don’t want to add these technologies just because they’re cool,” she says. “We want to add them because they will improve our work and our lives.”

Everson says she’s looking at how automated billing and AI-enabled risk assessment on construction sites can help her and her partner do their work faster and better. She’s also interested in how AI can help with data after a project is completed, such as with post-occupancy evaluations. “AI could help us collect and analyze data regarding previous projects, even when we’re on to the next one,” she says. “It’ll allow us to stay engaged.”

David Bell of Washington, D.C.’s Bell Architects, which employs seven architects, specializes in historic preservation and sustainable design. As a result, Bell is interested in tools that can help design net-zero or net-positive energy buildings. “You have to optimize the building design for elements like orientation to the sun and wind and window-to-wall area ratio,” he notes. “AI could serve as a tool that creates a high-performing building in this regard, perhaps generating previously unrecognized forms that are optimized for passive strategies.”

Bell says that while his firm doesn’t have a lot of research and development dollars to spend on such technology, it does have clients who might be willing to invest. “Hence we’d be using it more on a project-by-project basis,” he says.

Bernstein urges solo practitioners and small firms to keep abreast of developments. “I would keep an eye on this very carefully and try to understand the slope of the curve so you can think about what it means for your business before it catches you,” he says. “Ignoring this as another piece of technology that you don’t understand or care about is a dangerous proposition.”

Luthra has a similar take: “People tend to approach AI from fear, but being afraid doesn’t get us anywhere,” she says. “The future is coming, no matter what.”