Would you trust ChatGPT to pick out your outfit?

After the public release of ChatGPT, Jiwon Hong had an idea.

Hong, who is the CEO of YesPlz, a startup that aims to use AI-powered visual search to improve how shoppers find products and retailers recommend them online, noticed that members of the Facebook community, which he is a part of, often require clothing. recommendations. They often ask questions like: “Can you recommend my vacation to Hawaii?” Or, “I’m meeting my boyfriend’s friends for the first time. what should I wear?’

“I thought that with ChatGPT and our image tagging, we could create this fashion stylist,” Hong said.

YesPlz’s new ChatGPT Fashion Stylist uses ChatGPT to generate clothing recommendations from user prompts and then YesPlz’s visual search capabilities to suggest relevant products from the retailer’s catalog. It’s currently in private beta and hasn’t been released publicly, but Hong said they demonstrated the product at the recent Shoptalk conference and had a lot of interest.

It’s not the only company interested in how AI can play stylist. Online retailer Ssense posted examples of ChatGPT making clothes out of their products on Instagram.

But how good can an AI be at putting together gear when it doesn’t really understand what it’s doing?

Generative AI like ChatGPT takes AI’s capabilities for pattern recognition and effectively runs it backwards, as technology analyst Benedict Evans recently described to BoF. By being fed enough examples of something, it can follow the example and create a new one. It doesn’t really understand the inputs or outputs, though, so it tends to give fake answers that just look real.

In fashion, it doesn’t understand why one pair of shoes works better with one outfit than another, as a stylist would, but it can follow the examples it’s trained on. It turns out that this is enough to create a functional stylist.

When a BoF employee told YesPlz’s ChatGPT Fashion Stylist that a formal event required a classy outfit, she offered three options: a modern look consisting of a jumpsuit in a bold color like red or emerald green with metallic heels; and then a “glitz and glam” look centered on a gold or silver sequin dress.

“Keep the rest of the outfit simple with minimal jewelery and classic pumps to ensure the dress is the focal point of your outfit,” it advises.

YesPlz product recommendations appear below the ChatGPT response. These generally correspond to the products it offers, but are limited to retailers in its catalog, which sometimes results in discrepancies between ChatGPT’s clothing offerings and the products displayed.

Playing around with the tool showed its potential and limits. It can offer styling advice that goes beyond accessories, like telling me to wear what I’m confident in when asked how to dress to impress my partner’s parents. But it also had a tendency to repeat ideas, just change the style or colors, and the options offered weren’t always the most imaginative. The Coachella outfit idea was spot on, which means you’ve probably seen it before, like a maxi dress with strappy sandals or a cropped top with jean shorts and combat boots.

Creativity, in fact, may be where the AI ​​stylist struggles, or at least needs help from a human entering the prompt. (The prompt you use is important. For example, ChatGPT’s fashion stylist used the default for womenswear if I didn’t specify menswear.)

ChatGPT Fashion Stylist offers three suggestions for text Coachella looks, including a maxi dress with strappy sandals, a crop top with denim shorts and combat boots, and a colorful top with a denim skirt and platform sandals.

In February, when Ssense asked ChatGPT to create an outfit using its brands, AI suggested a black turtleneck sweater from Acne, straight-leg Levi’s jeans, and a pair of white Common Projects sneakers. Some followers who commented on the post were not impressed with the basic look. In later posts, Ssense provided more guidance, such as telling ChatGPT which brands to choose from or asking for New York Fashion Week looks to ride the subway, meet friends, and attend fashion shows.

For now, the company seems content to just use ChatGPT for fun social content.

“While we are still in the very early stages of understanding how AI will change our industry, the reality is that tools like ChatGPT and [AI image generator] Midjourney has already permeated popular discourse,” said Tom Bettridge, head of creative and content at Ssense, in an emailed statement. “While we as a culture all play together with these new tools and their potential, there is plenty of room for brands to join this online conversation.”

ChatGPT’s standard ideas of what clothes should and shouldn’t work together can be overstated. In a recent episode of the Clothing Articles podcast, which investigated attempts to create a computerized wardrobe from the 1995 movie Clueless, host Avery Trufelman found a couple who preferred when a randomly generated computer program created their outfits. Chaos, Truffelman realized, might actually be a great stylist.

In addition to offering new ideas to clients, human stylists have skills such as figuring out what clothes are suitable for a client’s body type and staying on top of trends. That’s why ChatGPT is not yet ready to replace them.

But it gets more powerful. ChatGPT is the latest version of AI powered can suggest recipes based simply on an image of food in, say, an open refrigerator. One can imagine that one day it could also come in different outfits based solely on a photo of the clothes in the closet. The algorithm can also be coded to include a level of randomness that will generate more novel combinations.

However, the biggest impact of these abilities may not be on styling, but on how we shop for clothes online. Hong notes that thanks to the way search engines work, we’ve learned to use keywords like “square neck orange tops” to find what we want.

“With ChatGPT, it’s just ‘What should I wear today?’ We can speak naturally,” he said. “That’s the direction we’re going and that’s what I’m most excited about.”

With additional reporting by Sarah Elson.

Source link