We’ve watched as IBM’s Watson has transformed from the trivia king of Jeopardy to a creative mind, capable of crafting new recipes like a tasty BBQ sauce. In that time, the supercomputer has also shrunk from a collection of servers the size of a room to a lean piece of software living on hardware the size of a few pizza boxes. Now, in conjunction with Bon Appétit, IBM is giving you a recipe app called Chef Watson, a cloud portal through which you can hijack Watson as your own personal chef.
“Something we’ve started to explore is that Watson can help people be creative and come up with never before seen ideas,” explains Steve Abrams, Director, IBM Watson Group. “We’ve been using the culinary arts to explore that activity.”
Chef Watson’s spartan interface begins by asking the user for single-word components defining what they’d like to eat, and then it generates a list of options based upon those criteria. So to create a new recipe, you simply type the ingredients you’d like to use and not like to use (like peanut butter but not jelly), the kind of dish you’d like to eat (like paella or empanada), and the “style” you’d like to try (like Provencal or Portuguese). Then, Watson will produce a full, step-by-step recipe for you to try. Actually, it produces 100, ranging on a scale from “classic” to “experimental.”
To construct these dishes, Watson is using the same core food science and human psychology it used in previous cooking endeavors. What’s new is that Watson is now thinking through the filter of Bon Appétit’s 9,000 recipes. It’s actually learning from what Bon Appétit chefs have done successfully for decades, taking into account ingredient pairings, amounts of ingredients that are generally used, and the steps that are generally taken across various preparations.
And so while the old Watson cooking demos produced recipes that were mere ingredient lists, Chef Watson can create whole recipes that, in theory, should not only work, but exist as a sort of algorithmic version of Bon Appétit itself--a magazine gone sentient. Chef Watson’s software interaction empowers a fairly rich dialog between the home cook and the knowledge lurking inside almost 50 years of food publishing.
“If you look at our website, that’s what our core user likes to do. People already use a recipe and change things,”explains Bon Appétit Digital Director Stacey Rivera. “What Watson does is give you inspiration.”
Neither IBM nor Bon Appétit is claiming that Chef Watson can replace the judgment of an experienced home chef. In fact, both expect Watson to make mistakes, like misjudging a proportion, or telling you to steam when you should sauté. It’s why Watson “suggests” a recipe that you could riff off of (or fix if it turned out badly). Over time, home cooks will serve as a feedback loop for Chef Watson, letting the software know what worked and what didn’t. “Part of what we want to learn from this is the delta between what [Watson is] providing and what it takes to get to 100%” Abrams says. But no rush on that, IBM. I, for one, am just fine with a world in which I can still laugh at a naive computer making the occasional mistake.