Need a hand writing copy, driving business and spurring creativity on your team? The AI revolution is here. Here are all the ways it matters for designers.
For the past few years, Silicon Valley has felt as trend-driven as a Paris runway. First it was cryptomania and NFTs. For a full year, entrepreneurs and futurists were loudly proclaiming that the future was digital money and ape JPEGs. It hasn’t come to pass.
Then, the metaverse: a vision of a shared virtual reality apparently so enticing that Mark Zuckerberg renamed his company in its honor and has been spending $1 billion a month to try and make it happen. At press time, the real world remains undisrupted, and the metaverse mostly resembles a clunky, depressing version of The Sims.
Given all that, one could be forgiven for thinking that the hype around artificial intelligence points to another boom-and-bust fad. But unlike its predecessors, AI is not a speculative concept that requires leaps of imagination to understand—in fact, it’s fairly easy, and sometimes even enjoyable, to use. The technology comes with plenty of controversy, ranging from ethical questions about intellectual property to its occasional habit of simply making things up. But it’s very much here, and it’s already having an effect on the design industry.
It’s far too early to make sweeping judgments about whether AI will usher in a utopian age or just steal everyone’s jobs. But it is worth taking a look at what’s happened in the 12 short months since text- and image-generating tools like ChatGPT and Midjourney exploded into public consciousness. We’ve collected impressions from the industry: what’s good, what’s bad and—of course—what’s just plain strange.
Every day, new AI-powered products hit the market. Designers are finding clever ways to put them to work.
By far the most well-known AI tool to hit the market is ChatGPT, an example of what’s known as a Large Language Model, or LLM. The technology that powers LLMs is quite complex. What they do is simple: They write. That capability has proven to be a boon for designers who express themselves fluently in images but find it tedious to string sentences together. Even those who don’t mind writing are using ChatGPT to finesse their copy.
“It’s something I use weekly,” says Long Beach, California–based designer Shaun Crha of Wrensted Interiors. “I’m creating copy for a website or newsletter and asking ChatGPT, ‘Can you make this sound more elevated?’ Or I’ll type in everything I’m trying to say and ask ChatGPT to edit it down into a few key thoughts.”
Some designers are using AI tools to feed the social media beast and write Instagram or Pinterest captions through a slew of new dedicated copy-creation platforms. “I use an app called PocketAI for almost every written word, including social content, blogs and email templates,” says San Diego designer Rachel Moriarty. “I then personalize the output so it’s more my vibe. Writing is not one of my creative gifts. Now, I feel like I have a copywriter on staff and I’m the editor.”
Another school of tools has emerged over the past year alongside text-based programs like ChatGPT: AI that can produce stunning imagery with a simple text prompt. One in particular has emerged as a favorite for creatives: Midjourney. If you’ve seen an impossibly fantastical, shockingly photorealistic interior on Instagram recently and wondered where it came from, odds are that it was created with Midjourney.
Designers are finding intriguing ways to work with tools like Midjourney. For Moriarty, it’s a fun creative tool—something to play with and “kickstart my brain if I’m stuck on a concept.” Architects, too, seem enamored with this usage of the technology: The principal of Zaha Hadid Architects recently made waves for saying that “most” of the powerhouse firm’s projects incorporate AI as a brainstorming aide.
There are more practical, client-facing uses for AI image generators as well. In Charleston, South Carolina, designer Jenna Gaidusek uses Midjourney to help in the brainstorming phase of a project to get on the same page with clients. “I use Midjourney to create inspirational imagery that I’ll fine-tune and include with renderings alongside actual paint colors and materials,” she says. “It helps me point the client in the right direction without having to spend hours looking for inspirational imagery on Pinterest or Houzz.”
AI is often pitched in extremes: It’s either a humanity-saving miracle or an industry-
killing menace. But somewhat lost in that polarizing discussion is the fact that artificial intelligence tools are simply fun, or even a helpful educational tool for those interested in dipping their toes in design. Vanessa Edwards is a case in point.
After starting her career in public health, Edwards left the field to become a teacher, but she recently became intrigued by the idea of a third pivot—this time, into interior design. Playing with Midjourney became a way to explore her passion without quitting her day job. “I just started playing around and I got hooked,” she says. “I didn’t have the time to learn SketchUp, so AI was a great way to explore quickly.”
Edwards found herself immersed in Midjourney’s peculiarities, creating fantasy furniture and interiors inspired by everything from Mondrian paintings to simple Pantone color schemes. Finally, she bit the bullet and enrolled in an online interior design course. “Everything I’m learning as a student, I can explore with AI,” she says. “Color, styles, space planning—you can play with that stuff so quickly, and it helps me expand my creativity.”
Though she’s still enjoying the raw generative power of Midjourney, Edwards says she’s come to crave more precise control over her creations—and that AI may have mostly been useful as a gateway drug. “I’m getting to the point where I want to sit down with SketchUp and make a room look exactly how I want it,” she says. “Midjourney is definitely not the tool for that.”
AI, like all new technology, has the potential to shake up the existing order. Here’s what designers need to watch out for.
Much of the 20th century saw blue-collar manufacturing jobs threatened by the twin forces of automation and globalization. Now, with the AI revolution, it’s white-collar professionals on the chopping block, as AI tools threaten to replace individual skills—and in some cases, entire jobs and industries. Interior designers, sadly, aren’t exempt.
In the late 2010s, the industry saw a rush of e-design startups looking to capitalize on the inefficiencies of an old-school industry. Many of them went belly-up, thanks to a combination of the long road to profitability, investor fatigue and abject mismanagement. This time around, it’s AI-powered startups looking to get in on the action. Over the past six months, a rapidly growing coterie of companies with names like RoomGPT, InteriorAI, CollovGPT, ReImagine Home and HomeDesignAI (among many others) have sprung up, all offering design services powered by artificial intelligence.
These apps present an intriguing paradox. Many can produce seemingly miraculous results with very little user input. But they’re all prone to making strange mistakes—replacing windows and walls on a whim, for example. They also do not show customers real, shoppable furniture, and making precise changes (“move that sofa seven inches to the left and make it green, please”) is impossible without manual intervention.
Though most entrepreneurs have been building these design engines with homeowners as the target audience, they have quickly become popular with real estate brokers who want to virtually stage their properties, or interior designers who simply want something to show tire-kicking clients before going to the trouble of making a rendering. These tools seem to be most in demand from the professionals they’re supposed to be replacing.
“If potential clients request something and designers are not sure they’ll sign the deal, before investing a lot of time, they will do quick generations using our software and send it to the client for the initial concept,” says Denis Madroane, the co-founder of HomeDesignAI. “They’ll then move to the real design or staging.”
Madroane, like many of the founders behind this generation of would-be design disruptors, says the technology does not currently pose a risk to real-life designers. It will be a long time before AI can settle a dispute between a married couple at loggerheads over a sofa decision, or get the contractor to show up on time. But to produce better and better renderings with real, shoppable links? That seems inevitable. “There’s a lot of market demand, and this will surely happen,” says Madroane. “The most likely outcome is that generative AI will serve as an aid for most, if not all, industry professionals.”
Code of Conduct
Earlier this year, a minor kerfuffle broke out on a local Facebook group when a designer was outed for representing an AI-generated image as an example of their own work. The picture—an image of a bronze egg-shaped bathtub seemingly overlooking the iconic desert landscape of St. George, Utah—aroused suspicion because residential development isn’t allowed near the distinctive red rocks, so the view would be impossible. A social media dustup ensued. But that particular incident is less important than its broader implications: What does it mean for the design industry when anyone can use AI to generate convincing project imagery? Like all things AI, it’s complicated.
On the one hand, the ability for anyone to quickly generate a portfolio of AI imagery creates the risk of a market flooded with dilettantes and dabblers who devalue the design profession in the eyes of the public. On the other hand, up-and-coming designers without a wealthy clientele often need a way to showcase their creativity. In many ways, this debate is a supercharged version of the existing one around sharing other designers’ imagery on Instagram, or posting renderings instead of final project photos. Established designers tend to be skeptical of that kind of social media behavior; newer designers often rely on it.
The way forward is likely an informal code of ethics around AI-generated imagery. “The entire relationship between designers and clients is built on trust,” says Leslie Carothers, a consultant who works with designers and design-world brands on digital strategy. “I think it is incredibly important that all designers—if they are going to use AI imagery to attract potential clients—label the image itself as AI generated and mention it in the first two lines of any Instagram caption. AI imagery has to be used responsibly and transparently by designers so that the trust equation is never broken.”
Image-generating AI tools like Midjourney and Dall-E were trained by processing millions of pictures so that they can tell the difference between a cat and a dog—and between contemporary and traditional style in an interior. Startups don’t usually share exactly what data their algorithms “look” at, but many have been trained at least partially on copyrighted art, setting up a very 21st century intellectual property battle royale.
At the heart of the disagreement are two competing claims: Artists say that AI companies have stolen their work for commercial purposes without paying for it, and many are suing to prove that point. In turn, these companies often point out that their tools have only been trained on the data—
a process not dissimilar to a human artist learning to draw by studying existing works. After all, you can’t sue one artist simply for being influenced by another.
The issue is still working its way through the court system. But ultimately it may only be a minor bump in the road: Adobe has already debuted a generative AI tool, Firefly, that it says was trained on “copyright-safe” imagery, meaning imagery that either falls in the public domain, or had already been purchased as stock art by the company. (Whether the artists in question knew they were licensing their work for AI training is another story.)
Adobe’s approach may win out. The courts may also rule that training an algorithm on copyrighted imagery is acceptable. But no matter which program they choose, designers will have to get used to the fact that using an AI tool means working with technology that has been trained on other people’s creative work. That likely includes other designers: Popular generative AI tool Stable Diffusion’s data set included a large chunk of Pinterest.
Anyone who has spent time playing with AI tools knows how surreal they can be. This new technology is bound to have unexpected consequences on the design industry.
Entrepreneurs are hoping to build a tool that replicates what an interior designer can do. Is there any surprise that they’re attempting the same for product design as well? The past year has seen a flowering of new tools—and new uses for existing tools—that put industrial design power into the hands of anyone who has an idea.
In the world of textiles, there’s Fabric Genie, a tool developed by U.K.–based company The Millshop Online, which allows users to input prompts ranging from the simple (“blue floral toile”) to the esoteric (“Theresa May surrounded by cheese”) and get a working repeat. The Millshop offers customers affordable yardage of their AI-designed patterns, digitally printed on linen and cotton-linen blends.
Then there’s art. Gaidusek has taken to using Midjourney to create artwork that can be used to outfit renders. If her client falls in love with the AI-designed piece? No problem. Gaidusek uses another program to upscale the image to a hi-res file, which can then be printed, framed and hung in their home.
Finally, there’s the world of product design. Commonly available AI tools can’t produce working shop drawings (yet), but they can create inspiration images that others can work with. A host of startups are developing AI-designed furniture; Ikea has an AI-assisted flatpack sofa; and Carothers is working with manufacturer Thompson to produce a solid copper sink based on an AI-generated design she spun up using Midjourney, which is slated to debut at the Kitchen & Bath Industry Show in 2024. Those efforts represent the earliest stirrings of activity—expect much more to come.
At least so far, designers seem to mostly use text-generating AI to help write marketing copy and emails. But tools like ChatGPT are designed to do a whole lot more. True to its name, you can “chat” with ChatGPT about a variety of subjects, getting feedback on a business plan, pointers on a contract—or, as Long Beach, California–based designer Jon Udoff discovered, advice on how to win an argument with a city bureaucrat.
Udoff was working on an ADU project for a client in Long Beach—a new unit atop an existing garage. Though he had been mindful in his design, placing windows strategically to avoid too much direct exposure to nearby lots, a city planner rejected his drawings, citing privacy concerns.
“I had used ChatGPT for other things, like writing letters, but I thought I would try it for this,” says Udoff. “First I asked if Long Beach could impose design guidelines on ADUs, and it said yes. Then I asked if the city could impose design guidelines on ADUs if it didn’t have them for other structures in the same zone. It said no, and Long Beach does not have any restrictions like that. I also asked ChatGPT where in the legal code this was stated, and it gave me a direct citation.” Udoff quickly cross-referenced the actual law, and it looked like the AI had got it right.
“I sent what ChatGPT had told me directly to the city planner, she backed off, and the plan went through. It saved me hours of work,” says Udoff. “And no, I didn’t tell the city how I had gotten the information.”
Like many designers, Jim Cappuccino has an intake form on his website. Earlier this year, the Boston designer was reviewing a submission when he noticed something curious: In the “how did you hear about me” field, the would-be client had written a surprising answer—ChatGPT. “I thought, ‘OK, this is someone I’m going to have to call back right away,” recalls Cappuccino with a laugh.
To train ChatGPT, its makers “fed” an algorithm terabytes of text data, much of it pulled from the open internet. That includes everything from recipes to research papers to, yes, articles about interior design. As a result, the algorithm is capable of referencing real people and local businesses. “The client explained to me that he had asked ChatGPT, ‘What are some Boston-area design firms?’ and my name came up,” says Cappuccino. “It ultimately wasn’t a fit on budget or scope, but he taught me how to use ChatGPT, so I’m glad I took the call.”
This capability of ChatGPT—that it names names—is not without controversy. Because LLMs will occasionally “hallucinate” and simply invent accurate-sounding information, there have been documented instances of ChatGPT making up false and even harmful information about day-to-day professionals. A radio host is currently suing the chatbot’s maker, OpenAI, for defamation after it “told” a journalist that he had been sued for embezzlement.
OpenAI has already taken steps to curtail this kind of mistake, and it’s unlikely that clients will get wild lies if they ask ChatGPT about your services. It’s also unlikely that this kind of open-ended search will lead to a ton of new business for any one designer. But here and there, it might. “I know ChatGPT can make stuff up,” says Cappuccino. “But I like to joke that, since the client had typed in ‘great Boston designers’ and my name came up, it was telling the truth this time.”