Designers, take a deep breath. Really, really, AI is not on the verge of stealing clients. No matter how good an algorithm is at producing beautiful renderings, it’s safe until it can handle the nasty complexities of real people in the real world. When the day comes when I can zhuzh arrangements and deal with angry calls from contractors, I start to worry.)
but we that is We will see technology that brings AI-powered design to everyday consumers. In particular, it is deployed by retailers who want to sell their own products and by independent platforms who want to cut a variety of products. A lot of the technology on the market today is impressive, but not really that useful. is.
What would such an app look like? Imagine pulling out your phone, taking a picture of a room, and getting an almost instant AI-generated rendering of what the space looks like in Memphis Milano style. please try. Then we add some parameter changes that the program runs perfectly and after a few seconds it delivers a purchasable list of products. that is Killer design app. But for that to happen, his three fundamental AI challenges must be solved.
envelope problem
The first AI-powered interior design apps shared similar problems. Not only did they redesign the room, they ripped it apart. Algorithms move windows, demo walls, and in some cases ignore physics and gravity entirely.This is an envelope problem: AI really isn’t know The dimensions of the room you are asking to re-imagine.
Of the three big hurdles, this one is probably the easiest to solve. Over the past six months, AI has greatly improved its ability to preserve the underlying physical properties of the images it redesigns. Already many current AI design tools are able to stick to the basic shell of the room. If developers can find a way to combine data from smartphone depth-sensing lidar scanners, the envelope problem will be solved.
precision issues
If you want AI to generate an image of a serene coastal living room with cream bouclé sofas and blue accent decor, you’ll be fine. If you like the result but decide to replace the linen with a bouclé, that’s hard, the AI might overcorrect and come up with a completely new room. AI tools on the market today are not very good at making small, fine-tuned changes.
This quirk is a byproduct of how these systems are trained. By “looking” at millions of images and averaging them, you can generate an approximation of a particular style. But the algorithm doesn’t “know” that the sofa is a sofa.
There are ways around this challenge that developers building AI design tools are grappling with. But getting AI to make ultra-precise adjustments successfully is a huge challenge. Especially when it comes to subtle requests that humans often make (“I want this exact room, I want the pillows to be more vibrant, the curtains to be more sheer, but not too sheer”). that’s right.
Product issue
AI tools can create amazing interior design schemes, but even if you wanted to buy an entire AI room, none of them are realistic. This is a product issue. Artificial intelligence only produces approximations of real furniture and decorations, not the real thing.
There are some exceptions to that. Iconic pieces like the Barcelona chair are widely used in popular media, so the AI is trained by looking at millions of images, often replicating them exactly. But most of the time the AI is generating something like a mathematical average of his 150,000 chairs, not a specific piece you can buy with DWR.
Developers are also working on this issue. Attempted solutions may involve a bit of reverse engineering. another AI scans images and tries to identify similar ones. In other cases, solutions include training the AI only on a specific retailer’s inventory. Neither works perfectly. Of all the technical hurdles, though, solving product issues is probably the most lucrative, so expect full coverage to find answers. Where billions of dollars are at stake, there’s a way.
Homepage photo: ©Andrey Popov/Adobe Stock
