Cooking Like a Computer: The Albacore Tuna Pancake

IBM’s party for its glittering new headquarters in New York City last month featured a menu of chocolate burritos with edamame and apricots and cardamom-spiced Brussels sprouts. At the helm was culinary-school-trained Florian Pinel, but the recipes weren’t Pinel’s own—they were created by IBM’s new artificial-intelligence computer, Watson.

You may remember Watson as the winner of a 2011 Jeopardy championship round, asking questions from answers read to it in natural-language format, while not connected to the internet. Watson used only the data its researchers inputted into it—not just information, but also how questions and answers are formatted in the English language and how puns, homonyms, and slang are used—to create its own form of intelligence. Far more than a knowledge database, Watson showed itself to be a brain, of sorts, capable of processing data, forming hypotheses, and modifying its process based on past mistakes.

Those three capabilities in one machine set Watson ways apart from other computers. (The last characteristic is what will keep it from ever becoming completely human.)

Three years after winning Jeopardy, IBM built Watson its own home in New York City’s Silicon Alley. In a 12-story glass incubator designed by Fumikiho Maki, the opening party last month showcased Watson’s spectrum of talents. Dr. Watson absorbs medical research and individual patient symptoms to diagnose illnesses and, perhaps in the future, recommend treatments. Coach Watson may help process statistics to assist professional sports managers achieve goals. And Chef Watson analyzes recipes, trends, flavors, chemical compounds, and cultural and ethnic tastes to create truly new dishes.

I would argue that the world doesn’t really need another Jamaican chicken or lasagna recipe. If you need one, more than a million are at your fingertips with a split-second Google search. But Watson thinks like a chef, processing for example Jamaican food to find that it often uses spices, peanuts, mangos, and plantains and combining it with lasagna to find that dish has noodles and layers. Putting the two together creates a unique dish for Jamaican Lasagna, which is entirely new to me.

Photo of food truck.

Institute of Culinary Education chefs explain a kebab dish at the South by Southwest conference in Austin, Texas. Photograph courtesy IBM/Creative Commons 2.0

Watson’s palate crosses cultural boundaries. In western cuisine the more chemical compounds foods share, the more they are paired together but in eastern cuisine the opposite is true. As one of three components of its forthcoming app currently in Beta testing, Watson creates recipes based on which culture you want to favor (American, Chinese, etc.)

The app is simple: choose an ingredient, a dish you would like to make, and a cuisine style, and Watson shoots out 100 recipes. Come on over next summer for my Watson-recommended Fourth of July Albacore Tuna Pancake.

In this way, the app is addictive and I feel a little like a 7-year-old playing Mad Libs. (Ingredient? Brandy! Dish? Pizza! Style? Christmas! Christmas Brandy Pizza!) But the recipes themselves are solid—a cornmeal pancake topped with tuna tossed with lime juice, ginger, and tomatoes. Sort of brilliant. And, as Pinel notes, a great way to reduce food waste by finding creative ways to use up ingredients that are about to go bad in your refrigerator.

The app also assigns a predictive score for how the whole combination of ingredients will work together, from “Classic” to “Unique.” A Chef Watson cookbook is imminent, containing two years’ worth of Watson’s greatest hits based on Pinel’s and others’ work. But the app provides true personalized recipe development, 21st-century gene mapping for the cooking world. “Cookbooks for diabetics aren’t always appealing,” Pinel says, “ because they don’t satisfy personal cravings. Plus Watson gives the opportunity to never eat the same thing twice.”

What’s next for Watson? Food companies and grocery stores are interested, of course, in customizing the app’s capabilities to specific products and create new products more quickly. Pinel, a software engineer who was hired by IBM 15 years ago and worked in ecommerce and business process management while pursuing his culinary education on weekends, understands Watson’s business applications.

Pinel says the fragrance and fashion industries have taken notice too. “As long as you are creating products made of smaller components, you can discover interesting and surprising combinations of components that work.

“The computer is learning how to make something that has never been made,” Pinel notes, “and that has applications well beyond food.”

 

Watch as IBM Watson wins an episode of Jeopardy.