Is Generative AI the New Fashion-Tech Bubble?

It’s been just over a year since Kering introduced Madeline, a ChatGPT-powered shopping assistant that consumers can use to search for items and get product recommendations on KNXT, an e-commerce site that the luxury group has quietly used as a testing ground for digital innovations. .

At the time, KNXT promoted it on Twitter (since renamed X) as the end to endless scrolling in search of the perfect luxury items. The reality was less dramatic: Madeleine’s answers proved that Limited and robotic in early testing, suggesting products that weren’t always best suited for the occasion and speaking in what sounded like marketing copy. A notice on the site now states that it is under maintenance without mentioning a reopening date. Kering did not respond to requests for comment.

Madeleine appears to have met the same fate as some other AI-driven experiments that have hit the market as excitement over the capabilities of large language models grows following the release of ChatGPT in late 2022. The test has begun The technology needed to conjure up design concepts, create images for marketing campaigns, write product descriptions, and chat with customers. McKinsey estimated in 2023 that generative AI could add up to $275 billion to the operating profits of the fashion and luxury sectors in the next three to five years.

There is still a great deal of optimism about the potential of generative AI and billions in investment flowing into startups trying to make it happen. However, rapid adoption seems less certain. “While more analysis is needed, rapid advances in generative AI have not yet led to an explosion in AI use among businesses between September 2023 and February 2024,” the US Census Bureau noted in its report. March report.

It’s not just chatbots that face challenges. Levi said Fashion business In a statement, it has no plans to expand the scope of the pilot program that was announced last March Using AI generated models To increase the variety of models on its e-commerce site. It was one of the earliest and most ambitious uses of the technology at the time, but it faced a firestorm of critics who pointed out that human minorities actually struggled to get modeling jobs.

“We do not see this pilot program as a way to promote diversity or as a substitute for real action that must be taken to achieve our diversity, equity and inclusion goals, and it should not be portrayed as such,” the company said in a statement. Her statement.

The decision was probably motivated as much by the criticism the company received as by any technological shortcomings. But the limits of generative AI are becoming increasingly clear. More than a year after generative models first captured public attention, even the most advanced ones Fabricating factsMake Basic math errors And produce images with physical images or Historical fallacies.

They can be useful for certain jobs, but often “in the same way that it may sometimes be useful to delegate some tasks to an inexperienced and sometimes clumsy intern,” as Technology Observer Molly White wrote recently.

It is too early to declare generative AI a failure. But questions are growing about whether it can live up to the extraordinary expectations placed on it.

“AI will ultimately be transformative, but GenAI has a lot of technical issues, especially with reliability, and is unlikely to live up to the current hype,” said Gary Marcus, a prominent AI skeptic who recently wrote about the possibility of the bubble bursting. During the next 12 months, he said in an email. “It may be years or even decades before most of these promises are fulfilled.”

Hype cycle

Emerging technologies often follow a similar path — so similar that Gartner, a technology research and consulting firm, codified it in 1995 and called it the “hype cycle.”

A new innovation appears that generates a lot of excitement and publicity. Based on some notable successes, expectations become exaggerated and reach their peak. But when early experiments don’t work, there comes a period of disappointment. If all goes well, they will eventually rise again as next generations of technology emerge, returns become clearer and adoption increases, although there is no guarantee this will happen. The so-called magic mirrors in stores never work, and dreams of transformation may never come true.

“We have placed GenAI in our hype cycle for retail, which is absolutely at its peak,” said Sandeep Unni, a senior analyst in Gartner’s retail practice.

A chart showing the placement of different retail technologies on Gartner's hype cycle curve, which rises sharply before falling sharply and then gradually rising to a plateau.
“Hype Cycle” from Gartner Retail. (Gartner)

Some retailers had hoped that LLMs could revolutionize online shopping, for example, by allowing truly conversational chatbots that understood a customer’s queries and intentions as well as the context around them, such as when a wedding is formal and a picnic is not . They can then answer questions and recommend products.

There’s still a long way to go before most retailers want one of these chatbots, whose knowledge comes from public data scraped from the Internet. Amazon, which has invested billions in generative AI, has received middling reviews for its shopping bot that it began testing publicly in February, with references to Rufus. Washington Post Considering that “Mostly useless” and they say they don’t trust its recommendations. Amazon said it will continue to improve the robot, which is under development.

“The biggest thing we learned was that people want the reassurance of experts,” said Jake Stark, co-founder and CEO of Good Sort, a startup focused on AI shopping assistants that recently changed its approach.

The company, formerly called ShopWithAI, initially had an AI chatbot that recommended clothes based on different celebrities’ styles. This product was not scalable, Stark said. It still offers a choice of men’s fashion but has also expanded to include watches, where its AI uses writing from a panel of experts to shape its suggestions.

One promising application of generative AI in fashion is design, where realistic accuracy is not a major issue. This technology could allow designers to quickly come up with new ideas and even maintain their style by training the AI ​​on previous work. Designer Norma Kamali is fully interested in this use of artificial intelligence and is busy creating a system capable of it Help continue its legacy Once you move away from its brand. Startups are They race to build their own fashion tools From image-generating artificial intelligence models.

However, it remains an open question whether AI-based design will take hold on a large scale. Some designers may reject it because they feel, whether true or not, that it replaces or devalues ​​human creativity, a sentiment that consumers can share. Selkie brand already She faced backlash from customers To create images using artificial intelligence. There are unresolved questions about intellectual property issues as well. Revolve was an early adopter and It released a small batch designed with artificial intelligenceBut the company refused to disclose whether it will continue to use this technology.

Hilary Taymor, founder and creative director of the Collina Strada brand, quickly worked on creating the Midjourney images In its design process She continued to use it, despite not liking the developments in the latest versions of the tool. What drives the creative capabilities of artificial intelligence These are in many cases the same unexpected results Which poses problems when creating text. As developers work to reduce these hallucinations, they may also be stripping away the creative power of AI. AI may be better at reproducing the stereotypical idea of ​​a dress, but that’s not necessarily what the design company wants.

“I no longer find it as creatively stimulating as I used to, so what I do is I put it on an older version to continue using it,” Taymur said.

The future of generative artificial intelligence

It may be possible to overcome these issues eventually. AI systems can allow users to adjust what is referred to as “temperature” — essentially the amount of randomness in the output — which may let you decide how creative you want the AI ​​to be. As for the obstacles that chatbots face, such as product knowledge and hallucinations, where AI outputs skew false or absurd, retailers are addressing them using methods such as fine-tuning models through specialized training and a technique known as enhanced recall generation, which allows the bot to extract answers from a database. Separate cognitive data.

“How to create a customer experience with this, mitigate the downsides — like hallucinations — and then leverage the unique benefits, that’s where I see this technology going,” said Tian Su, vice president of personalization and recommendations at Zalando, where she works. On artificial intelligence applications.

Zalando introduced its own AI-based shopping assistant to select geographies last year. While Su admitted the technology is not perfect, she said it still adds value for users. The half million customers who have conversations with a bot may not always know which search keywords to use to find what they want, but through AI, they can narrow down results or discover new products. There is no other technology that allows you to have conversations like this simultaneously with every customer, Su said.

Four iPhone screens display different interactions with the Zalando AI assistant.
Zalando AI Assistant. (Zalando)

She added that existing features could also benefit from generative AI. Zalando is developing a search form where the user types into the search bar and the results they see are updated in real time. It’s like a chatbot that takes away all the chatting.

There are simple tasks that generative AI already seems capable of, such as writing basic product descriptions. Adobe recently introduced tools in Photoshop that allow users to fill space with created images and create backgrounds that can be used for marketing assets. Taymur said she regularly uses ChatGPT to write professional emails.

And technology continues to improve. Gartner expects generative AI to reach a “plateau of productivity,” where viable products exist with mainstream adoption, within about five years. Challenges of solving LSD, intellectual property issues, security and regulation could hamper it, Gartner’s Onni warned. Retail companies face the additional hurdle of finding talent to help them test, scale, and scale generative AI projects that provide real returns. But unlike a fantasy concept like the Metaverse, generative AI is actually an extension of AI more broadly, Onni said, and the value proposition there is more established.

There are real limitations that need to be overcome first, and there are no guarantees that they will be discovered. But a product doesn’t have to be revolutionary to be useful. Generative AI may or may not change online shopping or the ways brands design and create images. We can end up finding hidden uses in the background that develop what already exists, The way algorithms already have.

“I think there might be more excitement than there is right now,” Sue said, emphasizing that even her “maybe” comes with a question mark. “But there’s something real about it.”

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *

Related Articles

Back to top button

Adblock Detected