I am the director of visualization. When we landed on this idea, I was tasked with a number of things, the first and foremost being to lead the visualization efforts in our office. In an ideal world, that would mean I would bounce between every project, and whenever there was a heavy lift on some renderings, animation, or diagrams, I would have a role in helping shape what those looked like, to give some consistency to the office and teach younger staff.

As far as I know, it’s pretty unique. There are some firms who tackle all of the rendering and animation work in-house like we do. Smaller firms often can’t do that.

Yes, it depends on what the project demands. There’s a lot more animation these days, but still tons of still imagery. That’s the stuff I love. I think there’s a little bit more power and art in that kind of work. And then there’s VR stuff that we’re doing. Some firms have really latched onto it and find it to be a valuable tool. 

We have an AI task force at our office right now. It’s made up of an assortment of people across every area of expertise in our firm. It’s interesting to sit on that because everyone has a different fear. We’re all worried about something, but we’re all worried about a different something.

In the world of visualization, AI looks really promising and exciting. But a lot of people think you can just plug some words into an AI program and get an image out just as good as any architectural rendering. We haven’t found that to be true. There’s a huge gap between your prompt and the image that you get out in the end. You still have to start with a sketch that becomes a 3D architectural model; with a program like Revit, that then becomes your final image through a rendering process. Then there’s this 10% at the end—that’s where AI can do some pretty cool stuff to help you elevate the quality of your image, but you’re not altering the originally authored image. You’re making tweaks and adjustments to things that you would typically do in a program like Photoshop. AI allows us to explore a lot of those things in more efficient ways. 

There are almost certainly firms out there who are producing competition entries where they’re blue-skying the whole thing with AI. They don’t have ethical or moral objections to that. We do. We’re the authors of everything we create in our firm, from the drawing sets, to the animations, to the renderings. It’s all coming from human hands at this point—until the robots completely take over.

I graduated college in ’06. Computers were in every office I’ve ever worked in, always, from the day I started.

More or less. When I was in college, my freshman year, we weren’t required to have a computer. And then by my sophomore and junior year, it was a requirement of every student. So I started on a drafting board and all that good stuff. But once in the profession, everyone was on a computer. I wasn’t there for the inception of AutoCAD, but my father’s an architect, so I’ve been around the profession a little longer than a lot of folks. When AutoCAD arrived, there was a lot of talk about how this would be a revolutionary thing that would eliminate the need for draftsmen. What actually happened is drawing sets got bigger. The same thing happened with Revit. Supposedly, it was going to free up all this time. But, again, it just created more-complex models, more-complex details, far more drawings.

It’s a good question. The metaphor I always use are food stylists, who will take a stack of pancakes and instead of using maple syrup, they’ll use motor oil because it looks better in a photo, or they’ll take some strawberries and touch them up with lipstick to make them vibrant red. 

I think you’ve hit on the truth behind it, which is no matter how much lipstick you put on the strawberry, if the strawberry doesn’t look decent to begin with, you can’t ever get it to that high level. 

There’s an underlying reality that you can’t render your way out of a bad design; you can’t render your way around a poorly thought-out space. You can put a lot of gloss on the image and probably trick a few people, but if the quality isn’t there at the outset, it doesn’t matter. You can’t push it over the top through Photoshop, or push it over the top with a couple of AI filters. It has to start with some good bones, or you’re not going to get a resonant reaction from a client. Most can see through the BS.

Oh, I wish I hand drew more. Most of my hand drawing these days is drawing on top of a rough rendering. I spit it out and think my way through how I might want to crop the image, where I’m going to put trees, people, cars, to help frame the image and create a narrative story within the image. I might do all of that by hand on top of the image with a piece of trace paper.

It depends on the designer. There’s a few that absolutely hand draw 100% of the time. There’s one designer in our office, wildly talented, who starts with SketchUp. And he really thinks of SketchUp as digital drawing. For him, it’s not a tool that he’s using to think through the tectonics of a building. It’s a loose and fluid tool that he’s using to iterate quickly and think through the ideas in his head. To me that feels akin to sketching on a piece of paper.

We’re in the process of it right now. The first part of our office-wide AI guidelines were rolled out last week. But we’ve been kicking them around for several months now. The initial AI guidelines include some very generic things about making sure when we do go into a meeting, if we’re going to record folks, that they know there’s an AI program in there listening to them. 

There’s also concern now around AI programs that might help you touch up a rendering, because you’re uploading your image to their giant repository of images. For a lot of the projects we work on, if there’s an NDA in place, we can’t do that. So having things like that in place early are important. 

This is a question that we’ve talked about in our office: How do we want to position ourselves with respect to AI? Are we at the forefront of it? Early adopters? Or are we at the far other end of the spectrum, where we’re the bespoke knife maker who lives in a cabin in the woods and makes 24 knives a year, and says, “No, in our office, we don’t touch AI. We’re here making these incredibly unique, one-of-a-kind buildings. The quality you get from us, the final product, is so much superior to this AI slop.”

The boutique architecture firm, right? Somewhere in the middle is us. We’re not going to be the early adopters, and we’re too big to completely turn our back on it. When you’ve got 80-odd people, when you’re designing $150 million buildings in Revit, and there’s AI tools now integrated into that program, you’re going to start utilizing those tools when they make sense for a project. 

The reality of it is some of the most base-level work will get taken over by AI. That’s inevitable. The scenario that I’ve read about that sounds totally believable, and simultaneously completely sci-fi, is when we have a meeting like this on Teams, internal to our office, with a group of designers, maybe a couple of people in D.C. call in, a couple of people in Charlottesville, a consultant, and then we would invite an AI or two to that meeting, and they would sit there and listen. At some point during the meeting, you could task the AI to go find images of similar project types to share with the client at the meeting next week. So instead of an intern sitting there and Googling images or searching through images of our built work, AI can do that in minutes. It’s tasks like that that are going to start to fall to AI. 

Is it going to eliminate architects? I don’t think anytime soon. There’s still an artistry to what we do. That said, music scares me a little bit. You can look at AI-generated music, and you’re like, Damn, they kind of got it. So if they can get there that quickly with music, can they get there that quickly with design? I don’t know. Will there be a day when someone goes online and gives a prompt to an AI, an elementary school for X number of students on this site, and it generates a series of design options? And then you choose option 2 and build the school? Maybe. People do something very similar with houses. 

I think with the complexity of the projects we do, the complexity of what our clients are looking for now, it’s probably still going to take months of meetings to get all that information out of all those different user groups, synthesize it, and turn it into a successfully designed space that you can actually construct. It’s hard to sit down and, in a few minutes, feed all of the correct information into an AI to get back a project that’s actually going to work for all the folks who will live and work in that building every day. It takes a lot of human beings going back and forth to get there. I don’t know if the AI could just intuit that without the same amount of iterating back and forth. It seems like we as human beings in the profession remain a pretty vital part of that process.

Featured image: “This university dining hall, right on the edge of a dense deciduous forest provided an excellent opportunity to highlight the building’s connection to the natural landscape. By setting this image at night, the use of wood on the building’s interior is highlighted, strengthening its connection to nature.” All images courtesy of VMDO Architects.