Artificial intelligence, a kitchen hold-up?

Artificial intelligence is now on everyone's lips. People are talking about it at dinner parties, complaining about it, scolding it, being afraid of it, being enthusiastic about it. The culinary world is no exception. Some reject the very principle, some take a timid interest in it, while others embrace it with gusto. One example is the Institut Culinaire de France in Bordeaux, which in January organized the first AI-based cooking competition with its pastry students. While France is just beginning to take up the subject, the Spaniards have never been afraid to approach cooking through the prism of science. Whether at the Madrid Fusión congress, the Science &Cooking World Congress in Barcelona or Ferran Adrià's El Bulli foundation, the question is already on the table.

Quebec-born François Chartier, who has lived in Barcelona for many years and was crowned World's Best Sommelier in 1994, is a regular speaker at these major events. In 2009, he published Papilles et Molécules, a book in which he arranges food and wines sharing the same aromatic molecules into families, and formalizes harmonious ranges with which everyone can play. Explaining scientifically what many had already discovered empirically - the harmony between strawberry and pineapple, mint and sauvignon blanc, thyme and lamblamb, rosemary and muscatel... - he lays the foundations for research and is gradually building up a veritable database.

AI as sous-chef

Thanks to this database, he was contacted by Sony, a Japanese giant active in the fields of music, cinema and video games. Its new weapon is artificial intelligence," explains François Chartier. The group wants to use it in what it already knows - sound and visuals - and decides to add gastronomy, because it's becoming a key subject."Sony asked the sommelier to collaborate, and acquired part of his database to make it the core of its research. That's how François Chartier will find himself on stage for the first time in 2019, at the AI Sum Tokyo summit, to show how his science of molecular harmonies can, thanks to AI, develop new recipes and aromatic pairings. "Then, in 2020, at Madrid Fusión, we'll be making a recipe live for the first time, in dialogue with AI." Dialogue. That's the word François Chartier insists on. He sees AI as a phenomenally powerful search engine. "That's how I use ChatGPT. Remember when Google or Yahoo came along, we didn't know how to search. And if we don't know how to search, we take everything they give us, including mistakes! Right from the start, I said to Sony: if you want to be successful in gastronomy, you have to create a dialogue. Otherwise, chefs will back off. You have to succeed in creating a sous-chef, the one he cooks with. When a chef puts his nose to a lemon, then tastes it, very quickly what I call his psychic palate kicks in; he already has two or three recipe ideas. But taste is much more complex than that. It contains the chef's culture, his background, his knowledge, his likes and dislikes, what he wants to create: a dish à la Bocuse or like his grandmother? There are so many things going on in his head at the same time. That's what we need to find out. For the demonstration, we imagined a Japanese chef coming to Barcelona to cook. He says to the AI: I'd like to work with soy sauce. What do you suggest? It pulls out a list of ingredients. But the chef quickly says: no, I don't want to work with soy sauce any more, but a chocolate sauce. Give me the ingredients for chocolate, and Ruby chocolate in particular. That's how we built the recipe live, to show how, by dialoguing with the machine, we were able to challenge ourselves."

Sony is in no hurry, and has the means to achieve its ambition in the field of gastronomy: "to equip chefs with AI to unleash their creativity". Working with top chefs is a way of positioning itself, while waiting to reach the general public. For, in the not-too-distant future, it is likely that robotics will work alongside AI to help the amateur chef. The intelligent fridge already exists, and it's not hard to imagine that the next step will be to be able to prepare, or rather cook, dishes with what's inside. And this, taking into account our tastes, our nutritional needs, the climate, even the month and day of the year, and so on.

A management tool

If there's one area where artificial intelligence is already being applied, it's that of restaurant management, in particular procurement, in order to reduce waste and therefore costs. Numerous start-ups have already created tools, such as the Inpulse platform, born in 2021 from an observation: in many restaurants, management is still done in an archaic way. Managing inventory requires processing a large amount of data with a great deal of finesse," explains Brice Konda, co-founder of Inpulse. It's a question of bringing together within the platform all the data from the restaurant's network - the tour plans, the mercuriale (the market price of products) with the latest prices. with up-to-date prices, availability, recipe cards - as well as sales reports, so it's connected to the cash register. In this way, AI will be able to automate all tasks, monitor material costs in real time and, for each point of sale, forecast raw material consumption, make order recommendations to avoid shortages, but also losses."

At Inpulse, five people are dedicated to data science, creating self-learning algorithms. They are "fed" by two major data streams. The first, that of sales directly linked to the checkout, retrieved daily. The second, external data streams - 300 weather stations, calendar events, school vacations, religious celebrations, sporting events, etc. - are fed into the system. For the time being, this tool is only aimed at restaurant chains with a fixed menu, the only one capable of providing reliable data over time. For a restaurant that puts daily specials on the menu, or often changes its menu, it's certainly less relevant for the predictive part," concedes Brice Konda. By helping restaurateurs to maintain their margins and save time, our platform enables them to focus on their core business, which is to make a great offer and welcome their customers well."

Tailor-made meals

But what about the customer experience? In November 2023, Yv Corbeil, director and chief creative officer of Niji, demonstrated to 500 Relais&Châteaux members gathered for their annual convention how AI can create an ultra-personalized dining experience. The more information we have on people, on what they do, on their environment, the more specific it allows us to be, and therefore to produce hyper-personalization," he explains. But it goes even further, because AI is self-learning. We don't need to give it a lot of information, it just looks for it itself. Our demonstration was based on a single individual. We were able to generate a table and menu that made sense only for that person. The only point of entry was to give the person's business e-mail address. From there, the AI searched for any traces she might have left behind: reviews on Google or TripAdvisor, places where she had traveled, photos she had published, posts she had liked, etc. All this was aggregated to prepare a menu that made sense only for her. Everything was aggregated to prepare the elements of a meal made specifically for her. The ultimate haute gastronomie experience, and one that won't be the same the next day, because in the meantime that person will have moved on and met new people. I did it for myself: the AI determined that twice a year, I go to Saint-Rémy-de-Provence, so I probably have family there, see lavender and hear the cicadas sing. She created a lavender vacuum-packed duck for me. But she also detected that I always dress in black and that, this year, I've listened to Paint It Black 137 times, the original song and a bunch of covers. So she came up with a three-chocolate cylinder for dessert, with chocolate brushed onto the tablecloth, which you had to crack and use like a spoon."This is how AI could become a chef instead of a chef, needing only clerks in the kitchen to execute its recipes, recipes adapted to the needs of each customer. The only imperfection is that it has none. While it's very personalized, it's also very dehumanized," he agrees. There's no emotion, even though I'm convinced that imperfection is what makes things human and memorable."

Will it be a matter of bringing imperfection into the system to give it that extra soul? We're not there yet, because in the world of AI, imperfection means mistakes. It still generates what we call hallucinations, nonsense. For example, a fork with a spoon on the other side of the handle. So, either it's a brilliant invention, or it's just a big mistake. But that's for humans to judge. And that's why, when it comes to AI, humans must have the last word.