GBA Logo horizontal Facebook LinkedIn Email Pinterest Twitter X Instagram YouTube Icon Navigation Search Icon Main Search Icon Video Play Icon Plus Icon Minus Icon Picture icon Hamburger Icon Close Icon Sorted
Guest Blogs

When Buildings Design Themselves

Automation could revolutionize architecture — by eliminating architects

Our conventional understanding of architects and architecture could be preventing a more rapid shift to sustainable building practices.
Image Credit: Flickr

Seven years ago, in my then-column for Architect magazine, I wrote that computerized automation eventually could fulfill the ultimate aims of green building by achieving dramatically better performance. Now the same magazine has taken up the same topic in a couple of recent articles. In June, Daniel Davis declared that architecture can’t be completely automated because “it is — for now — impossible to get computers to think creatively.” Last month, Blaine Brownell echoed this sentiment, citing a new McKinsey report claiming that “creative tasks are largely immune from automation.” Yet the implications of automating creativity are much bigger than either author lets on.

“Robot replacement is just a matter of time,” wrote Kevin Kelly in Wired a few years ago. “It doesn’t matter if you are a doctor, lawyer, architect, reporter, or even programmer.” Robotic manufacturing and other advanced industrial technologies are familiar, but computers also have taken on many white collar tasks, including customer service, journalism, and web design.

Kelly says nothing more of architects, though automated processes already are changing the profession. Computational design and parametric modeling are routine in architectural practice now, but often they merely facilitate architects’ pursuit of exotic geometry. High-tech eye candy. What’s still relatively rare is employing advanced techniques to improve performance significantly, and what’s nearly unheard of is automating the creative process entirely.

This is how designers work: We study a variety of possibilities and choose the ones that work best or we like most. Automation potentially can improve every aspect of this process and become, in Kelly’s words, “better than human.”

Computers can study thousands of design variations

First, innovation involves generating a large number of ideas to find a very few remarkable ones. Computers can study thousands of variations in the time it takes a designer to look at dozens, discovering possibilities that might never occur to us. As Davis reports, Autodesk, which makes the most popular computer-aided design (CAD) tools for architects, is developing software that “learns the same way we do,” only faster, says the company’s chief technology officer, Jeff Kowalski. “This is the biggest, most fundamental change that I’ve ever seen coming our way.”

The trouble is, the way architects normally use computers is to enhance, refine, or document our ideas, not to generate new ideas altogether. As I wrote last month, the architect-as-artist is driven toward highly personalized visions, and we often sacrifice other priorities along the way. In other words, what we like isn’t always what works best, and this could be holding back the entire profession.

In Paris this month, world leaders pledged to shrink greenhouse gas emissions markedly, and they cannot accomplish this without the building sector, which is responsible for nearly half of energy and emissions in the U.S. alone. However, a typical “high-performance” building achieves fairly modest energy reduction — 25-35%, according to the U.S. Green Building Council and the American Institute of Architects. And those numbers have flat-lined in recent years, so the industry is stuck, it seems.

Yet the National Renewable Energy Laboratory calculates that adopting current best practices can nearly double that outcome, getting to 50-60% reduction, without any additional costs. Applying that to every building could cut the total annual U.S. emissions by a quarter, half the amount needed to stabilize the climate by 2020, according to estimates.

While the information needed for architects to raise the bar is readily available, most of us don’t use it, but it’s easily be automated: for example, the engineering firm Thornton Tomasetti has developed automated tools to optimize structures for cost and carbon footprint. With the stakes so high, and human architects not stepping up fast enough, maybe Kelly’s right that “robots will — and must — take our jobs.”

What about making beautiful structures?

What of beauty? Architecture isn’t strictly about saving money and resources, after all. As I write in my book, The Shape of Green: Aesthetics, Ecology, and Design (2012), a growing wealth of research is revealing how people respond to light, space, form, pattern, texture, and color, and much of this information could be automated during design.

“Beauty is merely a function of mathematical distances or ratios,” explains computer scientist Daniel Cohen-Or. He and a team invented a “beauty engine” that subtly improves photos — with an 80% success rate, according to their polling. The “computational aesthetics algorithm” CrowdBeauty, launched this year, mines millions of photos on Flickr to find overlooked images with exceptional composition, pattern, color, contrast, and brightness. As the MIT Technology Review put it in May, “These guys have taught a machine … to recognize beauty.”

I know of few places more gorgeous than an aspen grove in autumn, but there’s no “design” there — just genetic coding and environmental conditioning. Architects Benjamin Aranda and Chris Lasch have used automated techniques to emulate the growth patterns of nanostructures, and Portuguese artist Leonel Moura has applied artificial intelligence to generate architectural forms by mimicking the emergent behavior of ant colonies. With sufficient computational power and speed, buildings could evolve the way any living system does and make design cheaper, faster, smarter, more efficient, more sustainable, and more beautiful.

The naysayers are wrong

Naysayers are plenty. Last February in The New York Times, Nicholas Carr declared that “robots will always need us”: “We exaggerate the abilities of computers even as we give our own talents short shrift.” Architects agree: “Technology is important,” Jacques Herzog told Vanity Fair in 2010, “but computers cannot do anything without the assistance of the human brain.”

Yet, according to estimates, machines soon will exceed the computational abilities of the human brain — possibly in the next handful of years but certainly during this century. Just this month, Elon Musk launched the OpenAI project specifically to “surpass human intelligence.” In his book, The Glass Cage: Automation and Us, Carr himself confesses that just a few years before Google created a self-driving car, many experts thought it couldn’t happen.

Cognitive scientist Margaret Boden defines creativity as “the ability to come up with ideas or artifacts that are new, surprising, and valuable.” Because machines can demonstrate all three, Boden maintains that debates about creativity and computers really are disagreements about what we value. “To accept robot creations as artistic expression,” Moura tells me, “means to deny humans the exclusiveness of creativity, and many people are not willing to do this.”

Artificial creativity isn’t science fiction — it could be the future of architecture. The only thing holding it back is architects themselves. Can we get smarter about solving serious new challenges, or will we risk becoming obsolete?

Lance Hosey is chief sustainability officer with Perkins Eastman. This column was originally published at The Huffington Post and is used here with the author’s permission.


  1. morganparis | | #1

    I can't speak to the
    I can't speak to the potential for automation in the design of large complex buildings but I suspect it's immense. In my own small practice focused entirely on single family homes, remodels and additions I've observed that technology is making support staff less and less necessary. I've been in the business long enough that I can remember when even a small practice would have secretarial help to manage business correspondence and filing. The transition to email and digital files transitioned us out of that situation long ago. Then as recently as five or six years ago it was worth employing an assistant to measure and plot as-built conditions - laser tapes, iPads and incremental software improvements are now making that a thing of the past. What I'd like to see next is a little ticker in the corner of the screen which is continually updating cost and energy performance information as I move a wall or change out a window. But when it comes to actual design decisions, with its deeply personal client interactions, I don't see that coming to our little backwater any time soon.

    Finally I think what this article may be missing is that machines have different KINDS of computational power than we do. They have long been able to trounce us at chess but they can't play basketball worth a damn. Hell, they can't even walk convincingly. And if you don't think that calculating multiple trajectories of the ball and of one's own body and other players at phenomenal speed and in real time, while simultaneously paying attention to to the coach, to strategy and to inter-player logistics needs incomprehensible levels of computational power you just aren't paying attention. How this applies to architecture I have no idea, but I can't help but think that we'll not see our human selves and our particular human abilities cast on the scrap heap of irrelevance in the foreseeable future. Or at least not the athletes among us, I guess.

  2. user-1119494 | | #2

    Machines have more patience than architects...
    Most architects prefer clients that do not continually agonize and change their minds, but so many folks have trouble deciding. A computer with an expert architectural program, perhaps combined with a genetic algorithm to bring designs closer and closer to the client desires, might come up with truly innovative, creative, efficient, and personalized designs.

    Human architects might still be needed to hold hands and comfort the clients, or to seed the automated program with a few initial possibilities that would fit the site well, but the client could decide on what is most important to them without facing the high costs of asking for changes.

  3. STEPHEN SHEEHY | | #3

    The machine-human interface needs work
    Supposedly simple devices, like phones, TV remotes and home computers don't come with useful directions. There is no way that a normal person will interact with a CAD program and end up with anything useful. [After I started this comment, I got the dreaded "internet explorer has stopped working..." and what I had composed was gone.]

    We worked with architects in the design of our new house. Our first meeting was simply to decide if we wanted to work with them. Once that question was answered, we prepared a list of wants, don't wants, hopes and concerns. We also had a simple floor plan.

    By the simple process of asking and answering questions, our ideas ended up in an initial design that was continually refined as we thought through issues and our architects helped us see the implications of each choice. We met every two weeks or so. In between meetings, we sort of lived with the latest design. We'd often take out a tape measure or move the furniture in our existing house to tweak dimensions or clearances. We'd then e-mail or call the architects who could revise the design accordingly. The design software was obviously helpful in speeding up the process of making changes, but without the human interaction, I don't think we'd have ended up with what we wanted. By working closely with our design team, which early on included our contractor, we ended up in a house we love.

    Before I retired from my law practice, I experienced the same changes that James Morgan noted. The need for support staff largely disappeared. When I started in 1981, everyone needed a secretary to answer the phone, type and revise documents, etc. Research was slow and cumbersome. By the time I retired in 2014, technology had allowed me to be more productive without any support staff at all. Obviously, technology has replaced the need for humans to do tedious calculations or do endless revisions to drawings. But I don't see architects (or lawyers) being eliminated any time soon.

  4. DC_Eakin | | #4

    Great Article!
    Very thought provoking. I think that the largest hurtle will be general acceptance of performance-optimized designs over the status quo. I think this stems from the lack of experiencing living in a high-performance design. Most potential buyers only know what they've seen built by others before them (who did the same thing) or what a builder's sales rep suggests. The overwhelming responses from owners who have built high-performance homes definitely show that there is a better way. Maybe entrepreneurs need some spec homes that prospective clients can rent for a week or so.

  5. shtrum2 | | #5

    missing my gyrocopter
    A problem with defining 'beauty' is that it's interpreted differently by different people . . . a task not unlike trying to nail a handful of peanut butter to a tree. Individual responses to Michelangelo and Greco versus Kandinsky and Pollok may be all over the aesthetic map. And viewpoints depend on several factors; history, precedent, context, culture, etc..

    But it's always tickling when someone says that technology is on the verge of taking over creativity and genius. I remember as an architectural student the wonderful futuristic drawings of Hugh Ferriss, or the fantastical city concepts of FLW and Le Corbusier, replete with flying cars and gyrocopters. Or the great Everyman, George Jetson, when he uttered those famous words: ‘These 3-day work weeks are killing me!’

Log in or create an account to post a comment.



Recent Questions and Replies

  • |
  • |
  • |
  • |