This story is from our 10th issue, themed “HOLDOUTS”—about the things we hold onto, and hold out for. Read the editor’s note here.
Also in this issue: “You Have to Give Up What You Have to Get What You Need,” by Jesse Dorris; “Working with Stone, Ethan Stebbins Finds the Seam,” by Ben Dreith.
Hi there, housemates,
It’s been an interesting month in the culture war over AI. A few things I’ve been tracking:
— Substack embedding Pangram, an AI writing detection tool, into its content management system. Shots fired by Substack’s CEO, Chris Best: “We don’t want to wait until your Substack app turns into LinkedIn before we start to learn and make progress.”
— LinkedIn releasing a “Looks Like AI Slop” button
— Claude embedding a digital watermark on AI-generated text—a move the company says is to remain compliant with new transparency rules in the EU
— Online discourse around AI-generated fliers—I think we can all agree that brush script needs to take a pause? (FWIW, at the record store my husband and I run, we do not allow AI flyers to be posted—but will hand you a sheet of paper and a marker!)
— Brenda’s analysis of template slop
It’s hard to say if there’s genuine momentum here in the fight for human creativity; it’s possible these are just the headlines that rule my little internet fiefdom. I also don’t fully buy that tools like a “Looks Like Slop” button actually aspire to do anything besides train models to make more convincing slop. But it is vindicating to see companies and creators recognizing that slop proliferation seems to have run amok and that something is needed to rein it all in.
Where we as creatives draw the line around our own use of AI is the topic of today’s reportage from Elizabeth Fazzare. Speaking with designers at Laun, Charlap Hyman & Herrero, and Studio Ahead, she learns where they’re digging in—and where they’re holding out.
How much AI can you take?
’Til next week,
Lila
⌂ ⌂ ⌂
Like what you see here? Want to support independent design media? Wrong House pays its contributors, and your support goes directly toward producing more stories. Please consider becoming a paid subscriber.
To AI or Not to AI?
For some designers, it isn’t a question
Written by Elizabeth Fazzare
Architect Rachel Bullock and designer Molly Purnell gave artificial intelligence the old college try at their Los Angeles–based design studio Laun. “We’ve tried to use it a couple times to research property descriptions or specific zoning code items,” explains Purnell, who notes that she was a “very early adopter” of ChatGPT three years ago. “It was wrong every time.” Since then, they’ve abandoned the use of any large language model for their business.
Among its boosters, there is a common argument that the AI battle has already been won: Everyone is using it, and you’d better, too; if you don’t, you’re behind. In reality, some people are not—or at least, not as frequently as that fallacy might make it seem. Many creative talents, in particular, are keeping their distance.
Where many designers draw the line is in allowing AI to touch the creative process. To create a residential interior, Bullock and Purnell take an analog approach: a client questionnaire that covers necessities like the number of bedrooms and bathrooms as well as intangibles like light, scent, and textural preferences. From there, the six-person Laun team sits around the conference table to sketch out ideas on trace paper, which are eventually formalized in a Revit computer model. Mood and materials boards are gathered, and the studio requires that all design development meetings occur in person. “It’s a lot more nuanced than anything that you can get out of an algorithm,” says Purnell. “It’s about evoking a feeling.”
At Los Angeles–, Mexico City–, and New York–based architecture and design studio Charlap Hyman & Herrero, the team begins each project pulling inspiration images from a physical catalog they have amassed over 15 years of practice. “We create a Bible for the project with all these things we’re interested in, and that becomes our North Star as we’re designing,” describes architect and firm cofounder Andre Herrero. “Our value is our ideas and our ability to problem solve.”
Homan Rajai, cofounder of San Francisco–based design firm Studio Ahead, has also banned AI from the design process. Like Laun, the studio sees a gap between nuanced human understanding and the smooth consensus-chasing hum of the machine: “The creative process is really emotional. How do you tell the AI, ‘Make this feel safe?’”
Once a design is digitally modeled, however, some firms do take advantage of AI visualization tools. Herrero uses Google’s Nano Banana 2 to render material textures and get quicker sign-offs on the project vision. The younger designers in Rajai’s eight-person office introduced him to AI software Tripo, which allows them to pull the vintage art and furniture pieces they have sourced themselves directly into photorealistic renderings. Tradeoffs like this save time, allowing designers to focus more of their working hours on creativity itself, they say. However, they are proceeding with caution. “As a tool to help convey the creative process, AI is great, but not in actually creating the objects or spaces,” says Rajai, who also designs furniture.
In the design world, like all creative industries, AI can have devastating copyright implications. Developers scraping artistic content to fuel generative modeling software can violate intellectual property law. Bullock and Purnell ask their clients not to feed their mood boards into generative AI software to avoid their conceptual work becoming part of the training data. But Rajai argues there are even more mind-numbing dangers. “If society is like a child, the algorithm is feeding you junk food,” he says. Using it in excess “makes the mind of the designer or creative team dull. You lose the human touch or even the human thinking process.”
Because it is trained on existing project images, any AI-based design is mimicry, rather than original thought—and its responses often lack the contextual considerations that a human creator would factor in. That results in AI-generated design that “just doesn’t feel rooted in anything,” argues Rajai. “It takes away from the magic of bringing artists and artisans together to create a reaction no one could control. The creative part of me likes that unknown.”
While Herrero sees a future world in which generative AI could become accurate enough to complete a set of architectural construction documents, he believes that artistry will always hold value. Increased adoption of AI might add to that, he says, but could play out a few different ways. “It comes down to what we value: Do we value craft? Do we value human-made imperfections? One scenario is that our mistakes are going to be seen as really special because a human made them,” says the architect. “The other is that no one knows how to do anything by hand anymore and we will start paying $1 billion for a vintage chair because the knowledge of a human building a chair is gone.” Any design aficionado certainly wants to avoid the latter—implementing judicious use of AI can help. ⌂
Elizabeth Fazzare is a New York-based editor and journalist who covers architecture, design, culture, and travel.
FROM THE ARCHIVE
PAST ISSUES
ISSUE 01: FINE, I’LL DO IT MYSELF ⌂ ISSUE 02: NO THOUGHTS, JUST VIBES ⌂ ISSUE 03: DON’T SIT ON THAT! ⌂ ISSUE 04: HOLIDAY SPECIAL ⌂ ISSUE 05: SWEATIN’ THE OLDIES ⌂ ISSUE 06: KITCHEN SINK ⌂ ISSUE 07: HACK JOB ⌂ ISSUE 08: OUT TO PASTURE ⌂ THE SHOE REPORT: MILAN DESIGN WEEK ⌂ A NYCxDESIGN CHORUS ⌂ ISSUE 09: GETTING WARMER
even with the AI visualization tools, we can educate our clients as to why optimizing everything is a danger to great interiors. the best rooms were are constantly tweaked and adjusted using our intuition. presenting AI renderings is neither romantic, nor a guarantee that everything specified will work perfectly.