Creepily customized dishes and facial recognition: how might AI change fine dining?

In the world of ultra-fine dining, service must run like clockwork. A team of specialists work together to create a seamless experience for customers from the point of booking reservations to the time the check is paid. Host, server, food runner, sommelier and dining room manager attend to – and even anticipate – guests’ needs with unflinching poise. When it works well, customers feel cared for and pampered.

It’s time-consuming work to pay such attention to detail, and early advocates of artificial intelligence (AI) say that software could automize the most tedious parts of the job. Let workers focus on the food and service, they say. Others wonder if it will erode time-honed traditions in kitchens and dining rooms. So when considering the question of how AI might affect haute cuisine, it depends on who you ask.

“Restaurants, especially in fine dining, are typically resistant to technology,” said the chef Jenny Dorsey, head of Studio ATAO, a research and advocacy organization at the intersection of hospitality and social justice. “It’s a human-driven industry, so the actual value and usefulness of AI versus the difficulty of adoption is an equation that isn’t totally solved yet.”

But some restaurants have already started to adopt AI to track inventory more efficiently, forecast sales or perform other routine tasks.

“The food industry is so busy,” said Katrin Liivat, who created FoodDocs, an AI software program that helps restaurants monitor, log and automate their food safety compliance. The company claims its software decreases the time spent on food safety tasks by 20% – or roughly eight hours per week – by tracking data and automating reminders for food safety tasks and shelf-life. “AI makes it possible for staff to focus on the more human side of their business rather than being stuck in compliance tasks,” said Liivat.

One potential application for AI is to create hyper-customized menu and wine recommendations for customers – something that many high-end restaurants already do the old-fashioned way. At one of the restaurants where Dorsey formerly worked, an open concept kitchen where chefs could chat with customers, the doorman would research the guests and offer the chefs suggestions on conversation-starters based on their findings.

“We’d find out they just went to Shanghai and would ask them about their trip during service,” she said. “I can see that being incorporated into fine dining through AI. It’s another way to make the customer feel special.” Maybe the restaurant isn’t offering a fully AI-customized menu to an individual high-dollar guest, but “if you know they don’t like peas, you can omit the peas in a recipe without them ever having to ask.”

Richard Blais, the chef-owner of Four Flamingos in Orlando, Florida, and Key West agrees. “The ability for businesses to better understand their customers and anticipate their needs is something hoteliers and restaurateurs have been trying to perfect for ages,” he said.

But can this go too far? Imagine that a guest makes a reservation, then the AI booking software researches the guest’s social media feeds, dietary restrictions and preferences.

Eric Ripert, the chef and co-owner at New York City’s Le Bernardin, finds this hypothetical application of AI disturbing.

“That seems very invasive,” he said. “It’s also a lack of emotional connection with the client. It’s automatic and artificial. It doesn’t have the emotional sensor that humans have.”

Dorsey agrees that it’s invasive, but she thinks it’s entirely possible. “By this time, we’ve all bought into surveillance culture thanks to social media and smart devices,” she said.

Jay Hack is an attorney at Gallet Dreyer & Berkey LLP, which partly specializes in hospitality law. He speculates that facial recognition software used in other industries could find its way into restaurants.

Some speculate that facial recognition software may find its way into restaurants.

“Casinos have been using facial recognition for years,” he said. In those cases, security cameras in casinos use facial recognition to turn biometric data into a non-photographic digital “fingerprint”. They then use that data to root out theft or identify high-dollar players.

The possibilities of such use aren’t all benign. An AI-powered reservation platform shared between thousands of restaurants could also potentially discriminate against diners, offering some better tables or service than others, determining no-show possibilities and refusing reservations based on data from past restaurant visits.

“If the same model is applied in the restaurant business that banks use to deny credit,” Hack added, “that information about a guest could inevitably be used for discriminatory purposes.”

Replacing the chef-auteur?

Across industries, employees worry that AI might one day displace them. But few of the chefs who spoke with the Guardian thought AI would supplant them.

Richard Blais thought ChatGPT could function as an assistant manager or do a restaurant’s marketing. But it could also lighten the load of existing staff. He recently typed “send an apology letter to Mrs Smith for her overcooked steak and invite her back to the restaurant” into software. It spat out a perfect response in six seconds.

The tech entrepreneur Nikhil Abraham, founder of the culinary AI startup CloudChef, wants to monetize Michelin-starred recipes through AI. Chefs from all over the world can submit recipes to the platform that can be duplicated in restaurant kitchens anywhere in the world, then be made available for pick-up or delivery within hours. Current “chef creators” include Thomas Zacharias from Bombay Canteen; Srijith Gopinathan of the Michelin two-star restaurant Ettan in Palo Alto, California; and the Bengali food historian and chef Pritha Sen.

For his part, Le Bernardin’s Ripert doesn’t see robo food runners at his establishment any time soon – or ever. But he did recently purchase a domain name that would restrict anyone trying to use his name or likeness in an AI software. “I would hate to see AI Eric Ripert cooking somewhere, and it’s not me.”

Yet there’s nothing to stop AI from using accessible information to create an “Eric Ripert” recipe. Recipes, restaurant concepts, plating, wine lists and menus can’t be copyrighted.

After we spoke, I asked ChatGPT to create a recipe for a seafood dish inspired by Ripert’s culinary style. The result: a simple, classic preparation for skate wing meunière. ChatGPT even paired a crisp, mineral-forward chablis with the dish, saying: “Chef Ripert often recommends chablis.”

Or take Nobu Matsuhisa’s hamachi with ponzu-soy sauce, said Good Stuff Eatery’s Spike Mendelsohn. This dish can be found in various versions in restaurants all over the world. “You can claim the origin of a recipe but not ownership,” said Mendelsohn. “What sets your recipe apart from everyone else’s is creation. That’s what this business is about”: individuality, expertise and something distinct emanating from a chef’s imagination and their kitchen.

Still, Richard Blais wants to know the answers to the intellectual property questions. “From where and from whom is the AI gathering its information? I just asked my AI app for a recipe featuring dry-aged duck, cherries and parsnips. And in seven seconds, it delivered five recipes. Where did it get the recipes?”

It got them from the vast online stores of information, but that information was originally created by people. Regardless of reports about McDonald’s launching its first all-automated location and Wendy’s drive-thrus manned by AI chatbot, many dining chefs aren’t worrying about job loss.

“Cooking is something that human beings are uniquely good at,” said Abraham. He’s of the opinion that “AI will always be playing catch-up with human creativity, and the humans who are at the edge of creativity will always outsmart AI and have experiences that are more valuable.”

Dorsey believes there’s an argument to be made that people don’t visit restaurants for the kind of intense customization AI might be able to provide in the future. “They’re coming for the chef’s point of view,” she said. “But fine dining guests appreciate a little pandering.” Dorsey prefers to look at AI as a tool that chefs can use rather than something threatening. “A computer can hold a volume of information that the human brain cannot and can offer more choices faster,” she said. “It’s up to us to evaluate the choices the AI provides.”

Mendelsohn said AI’s possibilities left him “excited and kind of scared” because “I’m pro-humanity. But I’m also interested in technology and how it can enhance our lives. But robots and software can’t replicate what a chef does, even if you can codify a recipe. Like, an eggplant is smaller or larger than before, the fire is a smidge hotter than the last time I cooked.” When chefs create and execute a dish, they’re using all their senses, plus intuition. And as of now, AI doesn’t have senses of its own.

Blais summed it up: “AI cannot taste in the way a chef can. Even if you’re repeatedly using the same ingredient, let’s say a piece of fruit, AI cannot account for ripeness, sweetness, texture. But I’m too smart to say never. Or at least I think I am.”

Before our interview, Liivat asked ChatGPT whether AI would ever replace chefs. She laughed: “Of course it said yes.”

Read More

Holly Kapherr Alejos