Discover more from The Horizontal Fault
Architecture's Endgame? Or a new beginning?
This will kill that; a few notes on the promises and paradoxes of AI for architecture
I have been thinking a lot about the nature of architecture and architectural practice this week— mostly in light of how large language models (LLMs; Pre-trained Transformers like ChatGPT), generative, deep-learning text-to-image systems (such as Midjourney and Stable Diffusion) and text to three-dimensional modeling tools (such as 3DFY.ai)— will transform architecture.
Although there is considerable boosterism, conjecture, fear and skepticism about the impacts of AI on architecture, I am especially curious about these tools, in light of what it will mean to be a sole practitioner, like myself, in the midst of what appears to be a rapidly approaching new technological, social and economic horizon. And while I can't say that my outlook is entirely cherry, I can also see some positive changes coming to architectural practice, despite what some already describe as AI's job-killing potential.
Thanks for reading The Horizontal Fault! Subscribe for free to receive new posts and support my work.
Although understandably, outright dread about AI tends to focus on its weird design effects, misuses, and job-killing potential for architectural creativity and design, I suspect real, lasting impacts will be felt most significantly in how buildings are delivered, measured, and managed- in other words, on the professional practice of architecture.
Practicing how to Practice
Imagining new forms of architectural practice is not a topic I've come to recently.
For instance, I spoke to Archinect twice in the last decade about working across different scales of architectural practice, from very small to very big. I was interviewed in 2015 while Design Director at global engineering giant AECOM and later in 2017 when I decided to strike out on my own after a short-lived business partnership fizzled. I'll post those conversations here in the coming months with some built and unbuilt projects from each period of work with other ruminations about what I have learned in nearly 30 years of practice.
In the meantime, at the end of three decades of “practice,” I realize that how we practice and why we practice architecture will require more actual “practice” with new tools and maybe even a new idea about what constitutes an architectural practice altogether. What is an architectural practice anyhow? What is its essence? Is a practice defined by its people or unique expertise, or its brand? Is it the tools we use? Or is it a style of architecture, a firm's signature?
Of course, the answer is all of the above, and to be sure, companies and investors are betting that AI, new devices, and various apps may replace people, their skill sets, current tools and even generate new creative styles of architecture, for better or worse.
While driving people out of work to squeeze profits is morally reprehensible, does the AI-promised end of so much of what defines a day-to-day architectural practice— the endless scheduling and rescheduling, the pointless code-researching, the drafting and redrafting, the interminable Zoom meetings, the note-taking, and the following up and the circling back— so bad?
Right or wrong, if it is possible that so much of what architects do daily, their work habits and routines, are going away, what will happen to all the architects whose livelihoods depend on the enormous but very human inefficiencies built into the processes of seeking entitlements and permit approvals, creating budgets, redesigning details, specifying materials, and managing job sites, schedules, and clients?
In 2005 or 2006, I gave a talk titled Minor Practice at Materials & Applications (M&A) in Silverlake. Unfortunately, I don't have any record of the discussion, but from memory, what I was interested in then (as now) was the idea of proposing a countermodel to a major architecture practice: that is, the now de rigueur, signature-driven, auteur-focused architectural celebrity led studio (see: OMA) that has come to dominate the so-called alternative architecture space; the counterpoint to mainstream corporate business-driven architecture.
When I set out to open a small studio in 1994, there was still a real distinction between self-described alternative practices and professional offices. The former usually prided themselves on creating personal, artistic, and even visionary work, and the latter marketed themselves as successful business entities and well-equipped professional organizations.
There were and remain many offices that need not fit into these categories and often successfully practice either a specialized, an original, or local type of architecture. Still, for now, and without polarizing the conversation, I'd like to focus on either end of the "architectural practice spectrum."
So, to be a bit blunt about it, these days, it is pretty easy to have a lot of trouble distinguishing between the “alternative,” tech and culturally-savvy Elon Musks of the architecture world (BIG and others) and the mainstream, business-savvy Apples of the architecture world (Gensler, etc. ). That the distinction between the two has collapsed is less the fault of those specific architects than how global capitalism has herded architects, much like artists and musicians, towards an increasingly flattened, streamed, and evenly distributed cultural marketplace.
This flattening of differences and the self-similar nature of so-called alternative and professional practices opens architecture up to disruption. To be more brutal about this situation, the increasing lack of creative diversity and genuine differences between architectural practices will only aid AI’s takeover of the field.
The most dire metaphor or image I have of architectural practice today is a dimly lit, disappearing, melting landscape— like the Arctic Circle. All that is solid melts into unemployment. Or perhaps a dying star, still bright enough to see by but burning through what's left of its energy and throwing off its last remaining rays.
Maybe that's all too poetic. But a big part of the corny promise of AI, admittedly, is architecture's rebirth after its collapse. And that may seduce us too much. But setting those romantic ideas aside, what I'm most interested in thinking through now is teasing out the promises, pitfalls, and paradoxes of AI for architects.
A day can't go by without some tech guru or politician proclaiming artificial intelligence's radically disruptive (or potentially destructive) potential. Just last week, The New York Times trumpeted that AI Demand Lifts Nvidia Toward Trillion-Dollar Valuation only three days after announcing that generative artificial intelligence could eliminate tens of millions of jobs.
So, big surprise, The Times seems to think that AI will be a feast for some and famine for most—in other words, more of the same tragedy that globalization and hypercapitalism has been serving up since the 90s. The news cycles we are witnessing around the potential of this technology swing wildly between predictions of a great catastrophe and a cornucopia of human potential. That may be the paradox of artificial intelligence for architecture and other creative fields.
Architects, perhaps unremarkably, are seduced by and scared of the potential of AI to transform how they design and work. Patrik Schumacher of Zaha Hadid Architects (ZHA) has been at the forefront of a conversation about the impact of these technologies on ZHA’s work. He has gone as far as to suggest that artificial intelligence can perhaps recreate many of the firm’s signature interests, formal concerns, and gestures. Schumacher is a big promoter of the AI's near magical on-the-fly ability to examine "…light, shadow, geometry, coherency, the sense of gravity and order [in a way that is] potent..." and is going so far as to admit that ZHA is developing "most" projects using AI-generated images.
On the bust side of the argument is a designer named Sebastian Errazuriz. Without much evidence, he claims that fully "Ninety percent of architects will lose their jobs as artificial intelligence takes over the design process." With that said, he seems to be spending most of his time doing things like vandalizing an augmented-reality version of a Jeff Koons balloon dog sculpture, proposing a rocket pad to replace the fire-damaged Notre Dame cathedral, and designing 3D-printed shoes representing twelve of his last lovers, so it is not entirely clear that he's been studying the architectural employment market that carefully.
If anything, one thing that AI is already successfully generating is a sort of architectural Bonfire of the Vanities. To get a sense (and a good laugh) of how inane the conversation about the impact of AI on architecture will become, I asked ChaptGPT to quickly generate a debate between Errazuriz and Schumacher on the effects of AI on architecture. I will keep you from the details; you can run the prompt if you are bored enough this weekend.
For sure, AI is rapidly proving that architecture’s cultural gadflies and influencers might soon be out of a job. It doesn't take a human being to auto-generate thoughtless or off-the-cuff opinions about technologies we don't know what to do with yet. Thankfully, AI is here to spit out all the nonsense, gossip, and misinformation we want without dealing with the hassle of listening to natural or so-called real persons. Maybe the disclaimer, 'No Real Persons Involved' won't be so bad after all.
But to be serious, what is attractive about AI is not only its ability to eliminate the drudgery that most architects struggle with but how it can quicken and sharpen an architect's design instincts. What is very exciting is how AI can quickly absorb, reframe and re-synthesize architectural conversations and perhaps even architectural thinking, something that has been the domain of academics and theoreticians. Think I am kidding. Here is what may be an interview in reverse in which ChatGPT allegedly chats with academic architect Matias del Campo.
But are we getting ahead of ourselves? Does mimicry equal imagination? Is using data analysis and pattern recognition tools the same as applying human thought to a human problem? And isn't architecture ultimately that most human of physical concerns— habitation— and not just robotic tasking? Isn't the art of architecture something fundamentally tied to our biological being in the world, what Gaston Bachelard described in the Poetics of Space as the "...embodiment of everything that is considered convenient, comfortable, healthy, sound, desirable, by other people[?]"
There's so much about architecture or all human creativity for that matter that cannot be replicated now or ever. So it is not clear yet that artificial intelligence spells either the end of architecture or the beginning of a new sort of architecture. What if it is just the continuation of architecture by other means, and architects must only decide now how to evolve?
And how different, ultimately, for better or worse, will architecture enabled by AI be? Will it be worse than using something like Grammarly? Are you a better writer after Grammarly? Probably. Are you a lazier writer? Actually no. You are now a faster and more precise writer. At least minimally, you are becoming a cleaner writer as the tool teaches you not to repeat yourself, to consider other ways of phrasing a sentence, etc. Or maybe that is just the tech seducing that ancient pleasure center in our primitive brains. Pavlov much, AI?
The Book will kill the Edifice
In the hands of architects interested in improving themselves (and therefore their practices), these new technologies should help and not hinder that very human need for growth, refinement, and, dare I say it, beauty. But it is also equally valid that, per Victor Hugo, many architectural roles may be reconfigured or lost, and we may be nearing end of the road for certain practices and business models.
While it might be too early to claim that robots will entirely replace architects, a few developments caught my attention this week and last. They are significant and will potentially transform architectural production in ways that will turn the idea of how we practice upside down. One was an announcement by Tel Aviv / Houston-based tech start-up Swapp. The company's founders have raised 11 1/2 million dollars to develop an AI-driven software that they claim will be able to produce construction documents, at least for simple buildings, in minutes and not in weeks or months. How this will work has yet to be wholly revealed, but in principle, they will use artificial intelligence to scour 3-D models and drawing libraries to produce standardized construction plans and details. For certain types of basic repetitive building types such as industrial warehouses— but even dormitories and hotels— this may put many high-production, low-fee firms on the ropes.
The potential of artificial intelligence will not only extend to automating repetitive tasks. For example, time-consuming activities, such as zoning and code research requiring architects to possess almost lawyerly skills, may be handed off to LLMs for analysis and processing. Likewise, much of what exists in construction codes can be codified and algorithmically processed. To that end, Upcodes has added Co-Pilot, an assistant that applies AI to Building Code Research. Several other elementary apps, so-called building code computing apps, are already on the market that can read zoning information. Some can produce crude zoning-compliant building envelopes and site layouts in minutes. In parallel, LIDAR-enabled apps like Iguide can make highly accurate as-built plans on the fly, linking two-dimensional traditional measured drawings to three-dimensional reference points and 360° images.
Is any of this artful? Probably not. Is it time and money-saving? Clearly. Will some architects abuse ChatGPT and get bogus data or made-up building codes? Inevitably. Just ask the lawyer who tried to generate a court filing with the tool and got a bunch of fake, invented cases replete with case numbers, courtroom numbers, plaintiff and defendant names, and filing dates.
We can and should expect some chaos until the platforms are refined.
Eggs will be Broken
What is more interesting creatively, at least to me, is that much of the sophisticated visualization work, which comes on early in the design phases of a project and has gainfully employed thousands and thousands of rendering artists— the very work that many architects enjoy doing on the front end of every project— is now under threat of extinction too—or at least reinvention. This week Adobe promoted the introduction of AI to its popular Photoshop software. The 'Generative Fill' function will arrive in Photoshop in 'the second half of 2023.' Adobe claims introducing this tool will create a "co-pilot" for graphic designers and not eliminate jobs, but that's hard to believe. The laborious process of producing and altering photographs and making detailed renderings of buildings will become a thing of the past. Instead, these and other tools will likely put the work of expensive professional rendering houses directly in the hands of practitioners.
While it is easy to be terrified or over-excited about the potential of AI, it's better to be slightly sanguine about what will (or won't) change, at least in the foreseeable future. So much is still unknown. So much of what will change for architects and how AI will impact them will still depend on other "natural persons:" clients, communities, building safety inspectors, builders, bankers, and politicians. And, any number of the unpredictable, very human political variables that often either guarantee a successful or disastrous or just an average building can't be so easily quantified or predicted.
The automation of much of the busy work that architects do these days may be a thing of the past. As in the now-distant past, hopefully, architects may have less paperwork to do and more time for drawing, modeling and design. Many onerous project management tasks may be automated, and construction-focused site meetings may be conducted with tools that verify and adjust construction decisions in the field (and via the building's digital twin.) Standardizing construction documentation may help reduce legal malpractice risks for architects needing more technical insight or better field coordination. Intelligent clash detection tools may eliminate many of the misunderstandings that inevitably get between the architect and builder. AI-augmented AR tools may help to align the actual building under construction with its digital twin, speeding coordination.
But the most powerful, lasting, and positive impact AI might deliver to architects will be the leveling effect it may provide for smaller or younger practices.
At least in North America, it is the case that so many young architects and small practices are regularly shut out of competing for mid to large-scale projects because of the mythos of the big corporate firm that we still seem to worship in this country. For example, unlike in many other countries, there is a belief here in the United States, whether substantiated or not, that mature, established large firms are always more responsible when handling more significant projects, like bigger multifamily housing projects or hotels. Another equally persistent, deeply ingrained myth is that boutiques or young, small studios do better with "high-touch" smaller projects such as private houses and artistic work (museums and whatnot).
While it may be true that larger firms do possess more extensively vetted technical libraries, employ more experienced staff, and often can draw on years of construction experience or a lifetime of working with a particular set of complex codes— for instance, the intricate healthcare knowledge that is required to layout a hospital— it is precisely this data-driven and algorithmic knowledge that LLMs and generative AI threaten to swallow up whole.
Undeniably there is no better source of reassurance for a client than the wisdom that comes with human experience and wisdom. AI will never replace that sort of insight and understanding. But it does raise fundamental questions: what if a small group of experienced senior architects decide to leave their stressful and perhaps less than well-paid jobs in an anonymous global corporate firm and set out to offer the same services and knowledge but empower themselves with the tools that will make the old work of fifty people the new work of five or ten people? What if a few young, talented, and ambitious architect-builders under 30 set up a start-up to automate the design-build process? The examples I cite are not hypotheticals; they are happening.
In the best of all possible worlds, the disruption AI will deliver to architecture will end underpaying jobs while creating new models of practice that might break open the monopoly on clients and projects that larger global mega-firms have understandably guarded for decades.
What comes after AI is anybody's guess, but if the technology means more possibilities for younger and older independent architects to thrive in ways that are not only personally and creatively liberating but potentially more economically rewarding, all the better.
Or that is just a naive projection, and we'll all be living on UBIs while the robots keep marching toward their version of paradise.
Thanks for reading The Horizontal Fault! Subscribe for free to receive new posts and support my work.