While this blog usually focuses on community recreation center design, this particular post will take a higher-level approach — addressing how artificial intelligence is impacting the work of architects in all sectors.
As AI continues to change the way people do their jobs and live their lives, here is one question all architects are bound to face (if they haven’t already): “Can’t you just tell AI to make that change?”
I hate to be the bearer of bad news, but things don’t work that way. AI is revolutionizing many processes, but it has not yet made the human design of buildings an obsolete pursuit. On the contrary, AI has unleashed unrealistic, sometimes even ludicrous, expectations of architects.
What AI can effectively do is visualization. For example, we use the term “entourage” when we make renderings come to life with the addition of people, furniture, plants and other visual elements to provide realistic portrayals of a how a facility might look when it’s in use. But that’s not actually designing the building.
The AI embellishments in this rendering of what AI considers a community recreation center’s restroom make it look more like something found in a hotel.
Granted, we can ask AI to generate design ideas. The ability to quickly create amazing visualization is one of the most wonderful things about AI. However, it also creates the appetite for more renderings and more options for clients to choose from. That appetite can become insatiable and burn up the perceived time AI saves architects. Worse, it can create churn.
For example, all but maybe a half-dozen of 300 AI-developed design possibilities will be slop. Think AI renderings of humans with the incorrect number of fingers on their hands, but only for buildings: steps leading to nowhere, gravity-defying structures, material usage that makes no sense.
It’s critical when embarking on a building project to understand that humans are still necessary to figure out how to make buildings physically and safely work in the real world.
That said, whether we like it or not, AI is redefining what “rendering” means. It’s no longer simply a representation of an idea. Rather, it’s become an extremely literal, photographically realistic image that creates a client expectation that if somehow we, as architects, don’t deliver exactly what’s depicted, we’ve lied about our intentions. That’s not true at all, but it’s a real risk.
A related (and also unrealistic) expectation is that if we can create renderings of, say, a gymnasium, an aquatic center and a locker room, we can just press a few more buttons to create detailed renderings of every room. Spoiler alert: Even with AI, it takes more than a few attempts to get a rendering to a presentable state.
Same thing goes for applying materials and fixtures to every surface and element depicted in a rendering. Design models don’t include every little detail, which would be a huge waste of time. AI fills in the gaps. For example, an AI-generated faucet in a locker room could look like the most beautiful faucet anyone’s ever seen — and one that does not actually exist, because AI generated it based on all of the faucets it’s ever been fed.
The attached rendering of a restroom features AI-embellishments that look fantastic. But as architects, we wouldn’t use those faucets — or backlit mirrors, for that matter — in a typical recreation center. In fact, the municipal client for this project told us, “This looks like a hotel.” Getting AI to render exactly what we want is still very difficult, and its embellishments stray from intentions.
Simply put, AI complements the architectural design and rendering process by adding context with ultra-detailed imagery. It inhibits that process, however, by creating expectations that are not practical or even achievable.
The efficiency gains we are seeing with AI are in the same areas as most other businesses, such as automatic note-taking, drafting language and data crunching. But most architects’ hours are spent in production, where AI is neither wholly effective nor trustworthy.
That is changing; the latest update of AutoDesk’s Revit (the dominant 3D design software architects and engineers use) includes an AI assistant to speed up some of the more mundane and repetitive tasks when designing renderings. And while I was writing this post, Claude AI announced new “connectors” to creative tools that architects use every day. That’s how fast AI is moving in this space; significant changes are happening as I type.
All that said, we still are a long way from trusting the technology with any heavy lifting.
Just 5 years ago, we architects thought we would be designing with goggles by walking around in our own projects via virtual reality technology. How wrong were we? Today, nobody is working that way; tech dollars are being spent on AI development instead.
No one knows what the next few years will actually bring; heck, this post could be obsolete in a few months. For now, though, AI does not replace the architectural design process.
At least not yet.
Editor's note: This blog was entirely written by a human.
