 |
|
|
|||||
A cozy, neutral sameness defines our era of interior design. Velvet sofas. Bouclé armchairs. All-white living rooms. Beds layered with fluffy faux-fur blankets. Calming sage green kitchen cabinets. You see it in furniture catalogs, social media feeds, perhaps even your own home. And we’ve got algorithms to thank. A decade ago, social platforms shifted from chronological feeds to algorithmic ones, optimized to show users what they were most likely to engage with. As many cultural critics have pointed out, those systems reward what is broadly appealing and shareable. In interiors, that has meant rooms that are soothing and inoffensivebut largely devoid of personality. “Algorithms are a mathematical equation based on the statistical middle,” says Christiane Robbins, a founding partner of architectural firm MAP Studio, who has studied algorithms’ influence on design. “Over time, the middle becomes what everybody thinks they want.” [Photo: Lulu & Georgia] Over time, algorithmic aesthetics begin to feel familiar, then comfortable, then indistinguishable from your own taste. Its subtle, says Sara Sugarman, founder and CEO of Lulu and Georgia, a furniture brand that she launched in 2012, just before algorithms reshaped the internet. Your personal style is influenced by these trends whether you realize it or not. You might decide you like a shade of gray without realizing its because youve seen it hundreds of times. But experts like Katherine Lambert, Robbins’ business partner, believe that change is coming. Consumers are getting tired of the visual sameness all around them. Home brands are realizing that they no longer have a distinct point of view that sets them apart from competitors. “We’re seeing a ‘design resistance’ emerging,” says Lambert. “Designers are rebelling against the algorithm.” Sugarman considers herself a member of this resistance. At Lulu and Georgia, she’s pushing back against algorithm-inspired design across her business. Instead, she’s empowering designers who have a strong point of view to create idiosyncratic pieces that draw the customer in. The majority of the brand’s revenue comes from products that it designs and manufactures itself, allowing it to create an aesthetic that stands out from other brands. [Photo: Lulu & Georgia] This strategy has been good for Lulu and Georgia’s bottom line. The company, which is self-funded and profitable, has been growing at a rate of 30% year over year for the past few years. And customers tend to be loyal, with a repeat rate of more than 50%, which is roughly double the industry standard. Lulu and Georgia offers a glimpse into how the world of mass-market interior design might be changing, as consumers want to break free from AI-generated sameness. The Democratization of Design Sugarman grew up immersed in design. Her grandfather, Louis Sugarman, founded Decorative Carpets in West Hollywood in 1955, catering to elite interior designers. As a child, she spent time in the showroom watching designers create custom pieces for wealthy clients. It was a closed system, where professionals controlled access and defined taste. That began to change in the 2000s, as the internet and social media gave a broader audience access to design inspiration. Mass retailers like Target, Ikea, and Wayfair made it possible to recreate high-end looks at lower prices. Sugarman didnt see this shift as a threat. It was incredible, she says. Design became more accessible, and it helped the industry overall. [Photo: Lulu & Georgia] She launched Lulu and Georgia as a digitally native rug brand before expanding into furniture and decor. But as platforms like Instagram, Pinterest, and later TikTok came to dominate visual culture, Sugarman noticed customers arriving with increasingly fixed ideas of what they wantedlabels like modern, coastal, or traditional that all pointed toward the same neutral, minimalist end point. For Robbins, this convergence makes sense. The rise of algorithmic feeds coincided with years of global upheavalfrom the pandemic to political instability. In uncertain times, people gravitate toward what feels familiar, she says. Sameness offers a subliminal sense of security. Algorithmic Design is Good for Business For home brands, flattened taste is operationally convenient. When consumers want the same sofas, colors, and textures, demand becomes easier to forecast and inventory risk shrinks. Searches for white sofas and bouclé furniture have steadily increased over the past decade, making those products reliable bets. If your business depends on scale and predictability, algorithmic sameness is incredibly efficient, Robbins says. You can optimize your supply chain, minimize risk, and flood the zone with products. [Photo: Lulu & Georgia] But Lambert is seeing signs of fatigue in her conversations with designers and clients. People sense that something is off, even if they cant articulate it yet, she says. Especially in [hotels and restaurants], everything looks interchangeable. Theres a global scroll nowwhere everything looks the same no matter where you are. In response, Sugarman has deliberately pushed back against algorithmic design. Lulu and Georgia does not use any trend-forecasting firms and resists letting past sales data dictate future products. This sets it apart from other furniture retailers. The forecasting agency WGSN has a robust interior design division which many manufacturers and brands (like LG and Knoll) use to decide what to make. Target, for its part, has built its own generative AI-powered forecasting platform called Target Trend Brain. By contrast, Sugarman empowers designers with distinct points of view to create pieces that dont yet exist in the market. Roughly 55% of the company’s revenue comes from products that it has designed and manufactured itself; the remaining 45% comes from products it has curated from other suppliers whose aesthetic fits in to Lulu and Georgia’s. The strategy is bearing fruit. Many of the designer collaborations sell out within days. Some of Lulu & Georgia’s bestsellers over the last few years look very different from the soft neutral styles that dominates our feeds: A red marble dining table with rounded leg, a wooden dining table with perforated holes on the base, dining chairs with unusual shapes cut out on the back. The brand collaborates with interdisciplinary designers including ceramicist Lalese Stamp, architect Ginny Macdonald, lighting designer Eny Lee Parker, textile designer Élan Byrd, and fashion designer Carly Cushnie, encouraging them to design what they genuinely want in their own homeseven if it means making a objects with no track record of selling. Products are often manufactured in small quantities to test demand. [Photo: Lulu & Georgia] One example is a small wooden vanity chair designed by longtime collaborator Sarah Sherman Samuel. Sugarman initially doubted it would sell. Most people dont have vanities anymore, she says. Still, they made a small run. The chair quickly sold out, with customers using it as a sculptural accent in living spaces. As with other furniture retailers, Lulu and Georgia also experiments with color through made-to-order pieces. A sofa designed by Macdonald is available in bold shades like mustard yellow and paprika red, produced only after a customer places an order. The approach allows the brand to test unconventional colors without overcommitting inventory. Sometimes, Sugarman says, those experiments become massive hits. [Photo: Lulu & Georgia] For Robbins and Lambert, this strategy works because it is rooted in specificity. Specificity is the secret sauce that throws off the algorithm, Lambert says. The more cultural, historical, and contextual knowledge you bring in, the harder it is for systems to flatten taste. As algorithmic sameness reaches its limits, they believe consumers will increasingly seek out brands willing to take risks. Were seeing fatigue percolate, Robbins says. I think were approaching a cultural tipping point. Designers who resist the algorithm are going to win.
Category:
E-Commerce
About a year ago, an advertisement caught the attention of Ashleigh Ruane, a PhD student in physics at the University of Cambridge. The ad was simple but unusual: Teach AI about physics. Curious, she clicked. She learned that experts across fieldsfrom physics and finance to healthcare and lawwere now being paid to help train AI models to think, reason, and problem-solve like domain specialists. She applied, was accepted, and now logs about 50 hours a week providing data for Mercor, a platform that connects AI labs with domain experts. Ruane is part of a fast-growing cohort of professionals who are shaping how AI models learn. According to Freelancer, thousands of new AI data training and annotation roles have appeared on their marketplace, with most of the growth taking hold in just the past 18 months. These roles range from highly technical expert tasks, like evaluating complex reasoning or diagnosing model errors, to nuanced judgment calls that large models still struggle with. Were entering a really interesting time period, says Freelancer CEO Matt Barrie. AI models need more and more data. Were seeing professionals from every field in every part of the world taking part in this AI data training work. The trend raises bigger questions: If AI models have already been trained on the open internet and vast corporate datasets, why do they still need human experts? What exactly are these experts doing? And how long will this new kind of work be around? AI has read the whole internet’and still needs real experts Theres a common assumption that todays largest AI models already know everything they need to know. After all, theyve been trained on millions of books, articles, papers, and posts. But industry leaders say domain experts are now more important than ever. Models trained on the entire internet can get you to an 80% answer, but in legal or tax, 80% isnt useful, explains Joel Hron, CTO of Thomson Reuters. Our customers demand a high level of accuracy and trust. Leveraging experts ensures accuracy to the highest degree that we can. Ana Price, vice president of supply at Prolific, which provides human data for AI labs, agrees that experts are becoming even more important as AI models move into regulated, high-stakes domains. The demand for human expertise and domain specific feedback from AI models is growing and growing and growing, says Price. As these models have gotten bigger, the errors are becoming harder to spot. Real expertise is needed to judge the substance of what models are producing, and not just the surface level correctness. In other words, the internet alone is not a substitute for structured professional knowledge. The more organizations rely on AI for serious, high-stakes work, the more they need experts to show models how real professionals think. What expert AI trainers actually do Linda Yu spent the last decade as an investor, deploying $4 billion of investments into technology enabled businesses. She started working with Mercor as an expert contributor a year ago, where typical projects involve coaching AI models to think like an investment professional. My role as a domain expert is to evaluate whether the model response is not just technically correct, but whether the complex reasoning behind the response is accurateincluding assumptions the model made, where it may have overreached, where it missed, and what a better answer would be, shares Yu. The work feels less like training an AI model, and more like mentoring a junior analyst. Experts like Yu say the work varies from project to project, and is being applied across industries from law, medicine, engineering, and beyond. Participants are typically paid hourly$85 per hour on averageand may be asked to evaluate a models reasoning on a technical question, rewrite incorrect answers into correct, step-by-step explanations, and compare multiple model outputs and choose which best reflects real-world practice. The output isnt generic content, but high-fidelity reasoning data designed to shape how AI systems operate. AI interviewers interviewing AI trainers The work requires real expertise, which means AI labs need data from experts who are vetted. To assist with the vetting, some platforms rely on AI interviewers to assess the actual expertise of potential AI trainers. Experts jump on a call, and they interview with AI, says Arsham Ghahramani, founder of Ribbon, an AI interviewer with more than 500 customers, including an AI training data provider who is interviewing more than 15,000 experts a month. Youll likely be asked the best interview questions youve ever been asked. AI interviewers assess experts for signals that would indicate red flags around expertise, like irregular response cadence, whether they respond naturally, and of course, whether they have the required expertise for a given domain. It was actually my first interview with not a real person, says Yu. It scanned my resume and came up with really relevant questions. After each answer, the AI interviewer acted like a real person and summarized what I said and asked a question that was a natural extension of our conversation topic. I was fascinated by the technology. AI now evaluates the humans teaching it, a reflection of just how far people have advanced model capabilities. The ‘last mile of information’ still belongs to humans One of the clearest explanations for why expert data remains essential comes from Mark Quinn, senior director of AI operations at Pearl and former head of Waymo engineering operations. He draws a connection between todays AI challenges and autonomous driving. At Waymo, we worked towards the last mile of autonomous mobility. Now, were working towards the last mile of information, Quinn says. Even though AI systems are being developed to close the last mile of information, the reality is that people may still prefer human expert validation if they need an answer on what to do if their dog ate some chocolate. The metaphor resonates across the industry. Even as models get smarter and larger, theres a world full of edge casessituations that require judgment, ethical reasoning, or domain-specific logic that isnt easily captured in general datasets. Some leaders believe the last mile will shrink but never disappear entirely. Hron of Thomson Reuters notes, The base models still have a long way to go to be truly deep. Expert systems and expert knowledge will help models climb to the next level. Price of Prolific adds, Weve only scratched the surface in terms of what AI can do. Humans are a critical piece of the puzzle, especially in niche domains. In other words, the future isnt about replacing experts. Its about scaling the expertise thats essential to making AI models better and safer. A new kind of knowledge work For Ruane, the physics PhD student, expert data work has become a significant source of income. She recently accepted a full-time position, but notes that her new job will only be 38 hours per weekleaving time to continue contributing to AI training projects. What shes experiencing is quickly becoming common: skilled professionals treating AI training work as a supplemental career path, flexible side hustle, or even full-time job. The work plays an increasingly central role in how AI systems operate. As models get more capable, the value of real-world expertise is being redefined, not diminished. Experts arent just using AI. Theyre teaching it how to reason, think, and act like an expert.
Category:
E-Commerce
Sitting on a coffee table in his Chelsea office in New York City and surrounded by framed wedding invitations on the walls, Justin McLeod is worrying about AI. Specifically, the cofounder and CEO of dating app Hinge is concerned that his usersmany of whom have asked him to their weddings over the yearsmight fall in love with it instead of one another. McLeod has spent the greater part of the past 15 years studying the dynamics of human relationships, including what makes one person fall for another, and he sees that chatbots offer exactly what many people crave. Why would I invest in these hard human relationships with people that are not always available or might reject me when I can talk to this thing that is right here and will always say the right thing? he wonders. On this sunny afternoon in late September, chatbots arent yet upending dating apps, but something sure is. Bumble, once the women-first darling, has shed 460,000 paying users since the end of 2024, prompting the return of founder Whitney Wolfe Herd in March. Shes embarked on an aggressive retrenchment campaign that has included laying off 30% of the staff. Tinder, meanwhile, has lost more than 1.5 million paying users since its peak in 2022. Its parent company, Match Group, has also recorded steady revenue declines for the past three years for its business unit that includes former stalwarts like Match.com and OkCupid. Match appointed Spencer Rascoff as a wartime CEO in February 2025; hes slashed head count by 13%. But one app in Match Groups portfolio stands out. Hinge, which has 15 million monthly active users, saw its paying users grow by 17% year over year to 1.87 million in the third quarter of this year. The app took in $550 million in revenue in 2024, and more than $500 million in the first nine months of 2025. Were the fastest-growingand, in fact, the only growingmajor dating app, McLeod says. (Thats not quite true: Grindr, with 1.3 million of what it calls average paying users, is also on the upswing.) Simply put, Hinge is crushing it, Rascoff said on Match Groups Q2 earnings call. Hinges competitors are facing problems of their own making. First was their aggressive pursuit of users, favoring quantity over quality, which has degraded the overall experience of many dating apps. Meanwhile, their lax policing of junk profiles and botsand simultaneous price increases for increasingly important featureshas forced users to pay ever more to find decent matches. People are just tired of endless, expensive swiping that doesnt convert into dates. And now a rising generation is emerging with an entirely different approach to dating than earlier users, putting apps that dont evolve at risk of being left behind. Gen Zs relationships are increasingly mediatedeven definedby screens. They still use dating apps, but theyre skeptical. Gen Z has set a higher bar, Match CFO Steven Bailey told attendees at Morgan Stanleys Technology, Media, and Telecom conference in March. They want [dating apps] to be safe, they want them to be effective, and they want them to drive the outcomes theyre looking for. But Hinge keeps growing because it has stuck to its promise that it succeeds only if users end up deleting it altogether. We want people to meet up and find love in person, McLeod says. That sounds obvious, but in the world of dating apps, it hasnt always been a priority. While other apps favored ease of use (all that endless swiping) over outcomes, McLeod remained relentlessly focused on designing ways to get his users off the app and dating, even if that meant inserting friction into the user experience. A lot of apps grew much faster than us because they were more engaging and exciting, McLeod admits. But he was playing the long game. McLeod is now preparing for the next stage of Hinge. The company has been rolling out a suite of AI-powered features to appeal to users with rustier social skills (ahem, Gen Z). McLeod is also taking his matching algorithm up a notch, extracting even more information from users to personalize and refine Hinges picks for them. To stop people from falling in love with chatbots, hes fighting AI with AIand trying to engineer something incontestably human: a messy, authentic love story. McLeod knows something about the complexities of the heart. He founded Hinge in 2011 while at Harvard Business School to help people find real-world connections. At the time, though, he was recovering from heartbreak. He had dated someone as an undergrad, but they broke up and got back together several times as he battled substance abuse issues. By the time he got out of rehab, she had moved on. Several years later, with Hinge starting to grow, McLeod conducted an interview with a New York Times reporter where he recounted the story of the one who got away. That inspired him to look up his lost love, who was living in Europe and engaged. Though they hadnt seen each other in nearly a decade, something sparked. She called off her wedding, and a few years later she and McLeod married. Hes recounted this story numerous times. It was even turned into a New York Times Modern Love column and then an episode of the Amazon show based on the column. But as polished as the anecdote is, theres a deeper truth within it: Vulnerability creates possibilities. A decade ago, when Tinder, Bumble, and other apps were orienting themselves around engagementmaking the user experience addictive but the outcomes questionableMcLeod mapped out a different strategy, aimed at fostering emotional risk-taking. He would require users to put in more work during the sign-up process and would place deliberate hurdles for them along the way, all in an effort to get them to open up, not just swipe. Jackie Jantos, chief marketing officer [Photo: Evelyn Freja] Today, Hinge requires users to upload a minimum of four photos and fill out at least three prompts about themselves. The process is designed toget users to slow down, think about what they really want, and present a more unfiltered profile. McLeod says the app tries to give users tasks that signal a level of intention and create a level of vulnerability so that you can actually create connection between two people. The longer sign-up process has made a difference: Hinge has found that users are 47% more likely to go on dates when they engage with the written answers on someones profile rather than simply the photos. Last year, Hinge introduced another hurdlea feature called Your Turn Limitsto curb ghosting. Now Hinge users with too many unanswered messages must send a reply or end the conversation before they can resume swiping. The company even gently nudges users into the real world: Its AI will invite users to set up a date if theyve been chatting online for a couple of weeks and seem compatible based on their conversations. Hinge also uses AI to scan the content of messages and deploys a notification to double-check with a user before they send a message that might not be well received. Thats all well and . . . millennial, but the apps newer challenge is helping Gen Z userswho make up 56% of Hinges overall user basefind value in the app. CMO Jackie Jantos sees a generation that was isolated during the formative years when relationships develop, and that often reverts to interacting on social media rather than in real life. Hinges Gen Z users tend to be uncomfortable with small talk and hyperfocused on digital body language, Jantos says. So theres a lot of reading into the speed [with which] someone replies, how long the persons message is, and what type of emojis and punctuation they use. Match CEO Rascoff puts it more succinctly: They have atrophied social skills and need more help showing up and connecting with other people. Hinges first feature for younger users, launched in 2021, was inspired by TikTok voice-overs. Instead of making users write out their responses to profile prompts, Hinge now allows them to record a 30-second audio introduction. It hit the sweet spot of willingness to do it if Im the person whos posting it and extremely informative if Im the person [experiencing] it, McLeod says. With more than one in five Gen Z adults identifying as LGBTQ, according to Gallup, Hinge has also given users an expanded menu of gender and sexuality identifiers to choose from as they set up their profiles. Gender, relationships, and relationship types are being redefined, Jantos says. In February, the company added Match Note, which allows users to privately share information with matches before chatting with them. People have used it to disclose their STI status or gender identity. (McLeod says single parents also use the feature to let matches know about their kids.) Hinge is tuning its marketing for Gen Z as well. The company has long featured real couples in its campaigns. But Hinge is now focused on stories that showcase all the intricacies and uncertainties of real relationships to show Gen Z that they dont have to be perfect. For 2024s No Ordinary Love campaign, Hinge enlisted writers like Roxane Gay and Hunter Harris to tell the nuanced, real-life stories of people who connected on the app, then published the essays in a zine. This year, Hinge followed up with a second collection, released as a printed book and on a dedicated Substacksupporting it with a flurry of ads in major cities. In one, a couple meets, hits it off, then breaks up for a few months before getting back together. Another tells the story of Lia and Ole, a couple fighting against their preconceived ideas of what they want from a relationship (Lia had imagined a romance with someone more established, more mature). Spoiler: Five years after their first datewhen they jointly deleted Hinge from their phonestheyre still together. Despite his fears of AI keeping Hinge users from meeting real people, McLeod is embracing it to help improve their prospects. In January, Hinge launched an AI-powered coaching tool to help users refine their profiles. Instead of just asking users to type in their response to a profile prompt, an AI chatbot can now interview the answer out of them. If a user says they like to travel, the chatbot might ask them for their best travel story to add to their profile. Those interviews serve an additional purpose: helping improve the apps matching algorithm. Until the advent of generative AI, Hinges algorithm primarily considered the profiles that users liked as they swiped and tried to surface similar onesbut it never really understood why a user might have certain preferences. Now, McLeod says, the algorithm can take in the content of a users profile to deliver better matches. Its thinking about what youve said, what theyve said, what your prompts say, what your photos are, and using it to predict whether you might like someone, he says. Its not waiting for you to send a whole lot of likes for us to learn your taste. If McLeod succeeds, he could lift the fortunes of Match beyond just Hinges revenue. Matchs data shows that Hinge subscribers already tend to use the app alongside one or more of the companys other apps. Rascoff now wants to encourage that behavior, letting users populate their profiles across other apps with one tap. From a financial standpoint, weve found that its additive, he says. The user spend on the second app does not detract from their spend on the first. Rascoff envisions that the matching algorithm behind these apps could also be standardized. We dont want to do AI stuff for the sake of it, McLeod insists. Even so, hes staking his apps future on it. McLeod anticipates that within five years Hinge will work more like a personal matchmaker. Users will spend less time on the platform sifting through profiles and sending messages. Instead, they might provide more information to the app on the front end and simply trust it to show them fewer, better matches. That would represent a sea change for the entire industry, McLeod says: Well think of swiping through endless profiles to find dates as a bit archaic.
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||