A few years ago, my professional career took an unexpected turn. In parallel with my engineering work in the Research Division of IBM, I had managed to get a Culinary Arts degree from the Institute of Culinary Education and work in restaurants on weekends, before starting a food blog. Cooking and computer science remained, however, two very separate parts of my life.
That changed in late 2011, when as part of a cognitive computing ideation challenge, a co-worker proposed a project that would mix computational creativity and food. Could computers help people be more creative by providing never-seen-before recipe ideas satisfying a few inputs, such as (to keep in line with a certain TV cooking challenge) a base ingredient? I didn't know where or how far the idea would go, but given my personal interests it sure sounded like an opportunity I couldn't pass up.
Three years later, this project, now called Chef Watson, is still alive and well. You might have read about the IBM Food Truck, the Chef's recent stints in the Bon Appétit kitchen, or how the system works. Even better, maybe you've tried the web app for yourself. But you may wonder, "Why does the world need such a tool?"
I partly addressed this question in a recent TED talk. That being TED, I assured the audience that cognitive cooking was nothing short of a solution to problems of World Hunger caliber. In the near future, Chef Watson will assist us in changing our eating habits to help prevent, or live with, conditions like heart disease, diabetes, or cancer, without giving up the foods we like, while also helping us reduce the ridiculous amount of food that we waste every year.
But on a more humble and pragmatic level, I think Chef Watson could be the first specimen of a new generation of smarter cookbooks. Looking at the history of cookbooks through the ages, it's amazing to see how little has changed.
Around the globe, the first cookbooks were meant to train professional cooks in the preparation of meals and banquets for the upper-classes. The earliest surviving example is Apicius, a.k.a. De Re Coquinaria, compiled around the fifth century AD. Most recipes were wrapped up in a couple of lines and did not include proportions. In the Middle Ages, proto-celebrity chefs collected their recipes for royals and noble families, first in manuscript (like Taillevent's Viandier), then in print (such as Le Cuisinier François by La Varenne). The recipes, though somewhat more elaborate, still consisted of succinct directions without any ingredient list.
The first real modern cookbook arrived in 1845: Modern Cookery for Private Families, by Eliza Acton. Not only was it aimed at the home cook rather than the professional chef, aligning it more with the ancient oral tradition of passing down recipes among the women of the middle and lower classes, it also listed ingredients with amounts, and provided cooking times for hundreds of simple recipes (rarely containing more than eight ingredients), plus techniques. Acton was a big influence on Mrs. Beeton's Book of Household Management, which followed a few years later and even more firmly crystalized the recipe format as we know it today: ingredients, method, cooking time, number of servings, images.
A century later, you could find cookbooks about pretty much anything, from National Tajik Sweets (Dushanbe, 1969) to Smoking Salmon & Trout (Madeira Park, British Columbia, 1982). Beginning in the 1960s, Nouvelle Cuisine refreshed the field by popularizing new techniques, inventive pairings, and recipes that diverged from classic dishes. The era of the celebrity chef was in full swing, with big names publishing tome after tome.
With the 21st century came a new revolution: the internet. Nowadays, anyone can post, search for, and comment on recipes online. But what's most striking, perhaps, is that this revolutionized next to nothing in the culinary letters, save a serious push for quantity to trump quality—I'm looking at you, cupcakes, or you, orange juice. In fact, while print publishing sees its revenue dropping regularly, the good old cookbook, now also available for express or instant delivery, either through your favorite e-retailer or in digital form, boasts steadily growing sales figures. The recipe format on- or offline hasn't changed much. At most, you can expect more ingredients, more instructions, more images, and some nice little intros providing a glimpse into other people's lives–the humble Tibetans, the petulant chef, the first-world-problem-ridden actor.
In my opinion, the internet's biggest contribution to the culinary world is essentially searchability. For the past 20 years, through commendable technical achievements, search engines have brought vast amounts of information in any domain imaginable to users' fingertips. One might argue, however, that the time has come to rethink this approach. Who really has the need for a tool that returns millions of results, for questions that can typically be answered in a few words? If searchability was once the best that technology could deliver, a smarter solution seems both desirable and achievable.
This is the promise of cognitive computing in a nutshell. Cognitive computing systems don't just ingest incredibly large amounts of data; they understand its content, summarize its meaning, and discover its underpinning relationships. They learn and interact naturally with people to help them make better, personalized decisions. And in the case of Chef Watson, the system engages in a dialog with cooks to create new recipes. These recipes aren't just random ideas to add to the online masses. Instead, they're made specifically for you, based on your inputs, your tastes, and your dietary constraints. And for those who think that size matters, the number of recipes you can make is nearly infinite. Now this is a cookbook worthy of the digital age, where dynamic content adapts to the consumer on the fly.
The idea of cognitive cooking is still its infancy—we really started focusing on a consumer application less than a year ago. But I want to imagine what the possibilities could be in the not-so-distant future. I have a corpus of recipes from sources that I trust, such as the Institute of Culinary Education and Bon Appétit magazine, Chef Watson's current partners. I haven't picked these recipes myself, but they're reliable, and their number and variety will provide the algorithms with the basic data they need. Then there are printed cookbooks that I own, most of which are now available in digital form. They may not provide as many recipes (although the 200-plus cookbooks on my shelves probably do!), but these are a better indicator of my unique tastes; i.e., I chose them specifically. These are the recipes that I actively want to try, or the ones I go to preferentially when I have something in mind. Chef Watson can digest these cookbooks, then turn out new recipes in the same style (with similar flavor pairings taken from the same set of ingredients), customized for my lifestyle, my preferences. I'm now able to endlessly extend or even mash up my cookbooks. Let the fun begin. Prepare Mario Batali's dishes with Bobby Flay's ingredients. Make your favorite celebrity chef's food without risking cardiac arrest or diabetes. Cook like the unlikely lovechild of Julia Child and Gordon Ramsay.
Cooking should and can be fun. Say no to powdered meal replacements.
DISCLAIMER: This post is in no way sponsored by IBM.