Watson is IBM’s computer that famously won at Jeopardy in 2011. The quiz show’s brain-teasers, rich with puns, homonyms, and slang, require more than a Google-like ability to spit out facts. Jeopardy demands creative thinking.
So does cooking. Watson’s creators believed it could think outside of the box to create unexpected and delicious flavor combinations, because it doesn’t have sorted “boxes” of knowledge like human brains do (unless, of course, Watson’s creators tell it to). Whereas a human brain might never imagine salmon and blackberries together, a computer armed with the best food and cooking information might.
Enter Bon Appetit, creator of tens of thousands of rigorously-tested recipes. As programmers inputted those recipes into Watson, the computer “learned” cooking techniques and—more important—flavor. In western traditions, for example, foods with similar chemical profiles are usually paired together; in the east, the opposite is true. (See Why Indian Cuisine Breaks All The Flavor Science Rules.) So if a user checks a box in the Watson app for “Asian” cuisine ideas (one of dozens of style options), the results give far more suggestions than simply adding soy sauce and ginger.
“It’s not about recipes, it’s about knowledge,” says Stacey Rivera, Bon Appetit’s Digital Director. “It’s really best for inspiration, for home cooks who want to be creative. It requires a user who has judgment and skills.” So if your ability is somewhere between beginner and Bobby Flay, Watson can save your time, your money, your waistline, and maybe even the tons of wasted food we throw into landfills each year, if you’re willing to alter your eating patterns a little (and shouldn’t we all be willing?).
Say you’re having a July cookout and need to make a healthy, inexpensive dessert for a crowd using the pounds of blueberries your kid just picked so they don’t go to waste. (Not that I’m currently in this situation.) Enter blueberries as the essential ingredient, click dessert and Watson recommends similar ingredients based on food pairing chemistry. You can adjust the four main ingredients depending on what is in your fridge or other preference. Come on over for my soon-to-be famous Black-Tea-Blueberry-Basil granita this weekend.
After you’ve narrowed preferences down, Watson recommends brand-new cooking ideas, based on recipes Watson has studied. And users can go more deeply into the app too, modifying Watson’s modifications. “This is how we satisfy a number of users, from the experienced chef to the home cook,” explains Florian Pinel, lead architect for the Chef Watson project for IBM. Its website calls the human-computer partnership a tool to “amplify human creativity.”
It’s actually a lot like working with other users’ comments at the end of an online recipe. A cook uses good judgment and makes the dish her own. And one user’s outcomes will never be the same as another’s because Watson “learns” and changes with every experience, much like a human brain.
As I wrote when I tested the app last year, it can be addictive and feel like Mad Libs—perhaps this is why my kid loves it so much. Name an ingredient! Name a course! Name what you don’t want! It suggested Roasted Blueberries with honey? That’s crazy.
But Rivera contends that even when Watson is “wrong” about food substance, it’s right about flavor. During beta testing, Watson suggested walnuts in a cocktail but the Bon Appetit team thought there was no possible way to do that. According to Rivera, a week later the team went to a bar and spotted a drink containing pecan syrup, confirming Watson’s nut notion. “Working with Watson has been an interesting process for us to think outside of our box and see food differently. It’s a creative tool.”
The app can be particularly useful for those with eating restrictions, who tend to make the same dishes over and over. It can be adjusted for almost anything—vegetarian, vegan, gluten-free, or just “no shellfish” or “no green beans” or “no tequila” (sigh).
Maybe coming up: meal planning. Boxed meal kits with pre-measured ingredients are trendy but pricey, and the environmental impact of shipping “local ingredients” across the country in trunks with ice packs is drives some people bananas. To replace those services, Florian imagines telling Watson “I’m a family of four, I want to make five meals this week, spend $10 per person and I don’t want red meat. Then have the ingredients for Watson meal suggestions—meals you choose—delivered to your home by a grocery store. We’ve not actively working on that but we’ve talked about it and we’ll see where it goes,” says Florian.
Watson’s cookbook was released earlier this year, with a history of the project and some recipes. But the Watson app goes beyond recipes to help up your cooking game. As Rivera says: “If you’re looking for a recipe app, this is not the place to go; if you want to expand your creativity, we are it. Nothing else like this exists.”
Blueberry-Basil-Black Tea Granita
Warning: This is not a tested recipe! It’s a general guideline given by the Watson app, and I’ve modified the master recipe from Bon Appetit to give an idea of what my results were. I’ll be using my creativity to tweak it.
1 1⁄4 teaspoons black tea
2 1⁄2 cups blueberries
2 1⁄2 cups chopped cranberries
3⁄4 teaspoon dried, chopped basil
1⁄2 teaspoon kosher salt
1⁄4 cup sugar
Stir sugar, salt, and 1/4 cup water in a small saucepan over medium heat until sugar dissolves and mixture just comes to a boil. Add basil and empty tea into saucepan; cover and let syrup steep for 30 minutes. Strain syrup through a coarse-mesh sieve into a small bowl. Place syrup and blueberry and cranberry in a food processor. Purée until smooth.
Transfer mixture to a shallow glass dish, cover, and freeze for 1 hour. Remove from freezer; using a fork, scrape mixture to break frozen portions into tiny pieces. Freeze, scraping mixture with a fork every hour, until mixture resembles fluffy shaved ice.
Divide among bowls. Serve with whole blueberries.
Active time: 15 minutes. Total: 4 hours