By Matt Gross
For nearly 20 years, I’ve observed the South by Southwest festival from afar, fascinated by the list of bands, movies, and tech companies coming every year to Austin, Texas—but frustrated beyond measure that I’ve never been able to go, catch a hot band on its way up, eat late-night migas, and feel like I was on the creative cutting edge. This past February, however, was particularly bittersweet. That’s because SXSW was host to something that seized my attention and wouldn’t let go: a food truck dishing up Austrian chocolate burritos and Belgian bacon pudding, recipes inspired not by an overly tattooed mad kitchen scientist but by Chef Watson, IBM’s cognitive computing system.
The more I read about the thought process behind Chef Watson, the more I wanted to install it in my own home kitchen. These recipes were the result of Watson “reading” thousands of recipes, and thereby coming to understand the deep connections between ingredients, not just in terms of how they’re used in recipes but on a fundamental chemical level. That something like a Thai-Swiss asparagus quiche could actually (reportedly) taste good was testament to the intelligence of the system (not to mention the skill of the trained cooks who converted Watson’s ingredient lists into tangible food).
In other words: Cool!
But not just cool, of course. At Bon Appétit, we like cool just fine, but cool and useful? That’s our sweet spot. And as I drafted an email to Steve Abrams, the director of the team responsible for Chef Watson, I tried to imagine how IBM’s system and Bon App’s might collaborate. The recipes we “fed” to Watson would have to be our own—they’d form the basis of how the system “thought” about food to discover new recipes. Whereas before the recipe pool was a hodgepodge—anything and everything that wound up on Wikia Recipes—now there’d be a curated collection of considered dishes, a food philosophy derived from the 9,000-ish recipes developed painstakingly by our test kitchen over the past decade or so. What could Chef Watson discover, I wondered, about food connections we’d been circling but never zeroed in on? What would it show us we didn’t know we already knew?
This is the key to Chef Watson, I think: the way it unlocks a chef’s creativity. The heart of the system may understand food based on Bon Appétit recipes and a library of chemical interactions, but when it comes to finally cooking that Austrian burrito, the tortilla is firmly in the hands of a human being. As Bon Appétit’s test kitchen discovered when it used Chef Watson to develop Fourth of July recipes, these ingredient lists really were inspirational, prodding senior food editor Dawn Perry to use a random thing like flour in a way she’d never expected.
In the few months since I first emailed Steve at IBM, I’m amazed at how far the collaboration has come. Since Dawn’s early-June foray into Watson-assisted cooking, the user interface has evolved to include suggested directions for preparing the ingredients, a leap that feels like science fiction—yet it’s already here, in a new version of Chef Watson that Bon Appétit and IBM have developed together.
If you’d like to participate in the beta of the new app, apply here.
While Chef Watson’s arrival has hardly slaked my Austin lust, once I start playing with the beta I’m going to have it gin me up a Mexican egg dish—migas!—those folks in the Texan capital can only dream of.
If you want to learn more about the new era of computing, read Smart Machines: IBM’s Watson and the Era of Cognitive Computing.