Ludwig, a new artificial intelligence tool, wants to decorate your home. But we found some of its choices perplexing.
“The Computer Research Institute of Montreal helped us develop the initial blueprint and how we went about the blueprint, and then from there we started hiring the right developers and the right machine-learning engineers,” says Fulhaus CEO Andria Santos, of how Ludwig came to be.
I happen to be decorating a new home, so I figured I’d give Ludwig a try myself. I used the free version, which allows you to design one room. To furnish an entire home, you’re prompted to pay a $99 fee. To start, I uploaded a few images from Pinterest — such as photos from store websites and design blogs — then selected which space I wanted to focus on. Ludwig gave me the options of dining room, living room or bedroom.
Once I selected a room, I was rewarded in seconds with several furniture options related to the image I uploaded. From there, I could refresh the design page for new options, upload another image, or skip right to buying and contact the Fulhaus team to order the pieces for me from its vendors, which include brands such as Four Hands and Loloi. I could also rate Ludwig’s recommendations on a scale of one to three smiley-face emoji; in response to a poor rating, the tool told me: “Sorry, Ludwig is still learning.”
Ludwig was fast — but ultimately, I didn’t feel prepared to furnish a whole space after using it. I found that the pieces the tool chose didn’t mirror the images I uploaded as precisely as I expected. For example, I assumed a photo of a white couch with a round coffee table and a few chairs would result in furniture options that mimicked those basic specifications. Instead, Ludwig sent back a gray sectional, rectangular coffee table, accent chair, accent tables and a rug. I didn’t find that the style of those pieces was particularly similar to the image I’d sent either, or that they went together all that well. Without a designer’s eye, I don’t think Ludwig delivered the context or consistency in style I would need to arrange the pieces myself.
I asked the team at Fulhaus about my experience. They said that because inspiration can come from so many sources, they built Ludwig to interpret images as a general direction, rather than something to copy literally. So, when the tool makes recommendations, it is suggesting a combination of pieces that reflect the look and feel of the inspiration image, not a replica. They also assured me that as Ludwig learns more, it will get better at identifying pieces and styles that match specific inputs.
Ludwig is intended for typical homeowners like me, but also for professionals. “It was important for us to create a tool that could be used by a non-designer, but powerful enough to be a useful tool for a professional designer as well,” says Santos. In particular, Santos emphasized Ludwig’s ability to quickly sift through choices from some 300 furniture suppliers: “There’s no way a designer can get through those 300 suppliers — especially when your client needs to go live with a property in a month.”
I asked a couple interior designers if they would give Ludwig a try, too. They reported similar experiences to mine.
“I tried to use it three different times [with images from] my own portfolio to watch what it spit out and it was not even measurably close,” says Kristen Forgione, principal designer and creative director at the Lifestyled Co. in Scottsdale, Ariz.
“All of the results I was looking at had about 20 percent of the suggestions work with the style and aesthetic of the image that I copied and pasted in,” echoes designer Phoenix Grey of Orion Studios in Toronto.
Okay, but isn’t that exactly what human interior designers would say about artificial intelligence that could theoretically threaten their livelihoods?
“I never thought for even point-one seconds about how AI would affect design and designers until I tried out Ludwig,” says Forgione. “I do think there’s going to be a customer for [AI design]. There’s always going to be that person that wants it done the cheapest and the fastest — even if the furniture pieces don’t look anything like the inspiration photo provided, they’ll buy it. But there’s no human or personal connection.”
Grey took a similar view. “When it comes to coding and doing anything numerical or mathematical, it’s kind of scary how accurate [AI] can be,” he says. “But when it comes to the creative fields like digital marketing or interior design, you need to have a sense of style.”
Indeed, even Fulhaus’s head of design, Pierce Atkinson, acknowledged the need for a human touch in the design process: “We’ll always need that creative mind.”
Ludwig also does not account for budget — which can mean users find themselves with a shopping list they can’t afford. For instance, it suggested that I buy a floor lamp that cost $1,368. “In one case, the total for six items was $13,000, which is definitely out of the grasp of the average person,” says Grey.
However, both he and Forgione agreed the tool could at least serve as a user-friendly starting point for someone trying to figure out their aesthetic. “It offers ease of access with dragging and dropping an image into it, and having it show you some actual products that could relate to a similar kind of style,” says Grey.
Even if Ludwig didn’t nail my design preferences, I, too, found the tool easy — and fun — to use. And not every piece was a miss — I liked an accent chair with rattan detail and a bleached wood coffee table. The experience felt like a more entertaining and slightly more focused version of online shopping, if not at all like working with an actual interior designer. But hey, “Ludwig is still learning.”
Jamie Killin is a writer in Arizona who covers business, hospitality and lifestyle.