In late January, Sourcebooks announced it would be publishing a cookbook by Watson, IBM’s supercomputer best known for beating human contestants on Jeopardy! The book, Cognitive Cooking with Chef Watson, is a collaboration between IBM and the Institute of Culinary Education, and is due out in April.
The culinary experts at ICE developed 65 original recipes, inspired by unconventional ingredient pairings generated by the supercomputer. Because a computer can not technically “taste” in the way a human can, Watson looks at ingredients partnered in existing recipes, according to Florian Pinel, senior technical member, IBM Watson.
“Starting in the food domain, we showed how Watson could ingest vast amounts of culinary data, understand content and flavor profiles; analyze incredibly large numbers of ingredient combinations; summarize their meaning; and discover their underpinning relationships; in this case, unique and tasteful ingredient pairings,” wrote Pinel in a blog post about the new book. Going forward, the hope is that IBM can use such “cognitive-based apps” to help inspire foodies, and find solutions for those with dietary restrictions.
Over at ICE, the chefs have a “hands-on,” interactive interface with Watson, enabling them to prompt and steer the system to some degree, said Michael Laiskonis, creative director at ICE. “Among the hundreds of outputs that the system creates, some may offer more surprise or greater appeal to the chef, which also adds an initial personal touch to the flavor combinations chosen.”
From there, it is then up to the chef to make decisions based on his or her own skills and sensibilities to translate that list of ingredients into a workable dish. “While there were many moments of surprise, I cannot think of any exercise that presented an unpleasant dish,” Laiskonis said.
One particularly notable dish, according to Laiskonis, is Watson’s Vietnamese-Apple-Kebab. “Among the ingredients chosen were chicken, pork, mushroom, pineapple, and strawberry—all included because they share high levels of the flavor compound gamma-dodecalactone,” said Laiskonis. “While strawberry may seem an outlier in this group based on our conventional food associations and cultural biases, the combination worked wonderfully.”
Another standout recipe was the Austrian-Chocolate-Burrito. “By setting parameters within the system to ensure that we received a few ingredients common to a burrito—tortilla, ground beef, and cheese—the remaining ingredients took the dish to another level entirely, namely the chocolate, apricot, cinnamon, and orange peel,” Laiskonis said. “While the basic structure of the final dish was instantly recognizable as a burrito, the final combination of textures and flavors were a surprising hit.”
When asked if those at ICE had any hesitation introducing a technological element to a field known for having such a human touch, Laisknonis said the human chefs had no doubts whatsoever. “At the very outset of an exercise with Watson the initial spark of inspiration comes from the chef, and after Watson crunches its extensive amounts of data to produce its output of suggestions, it is then again up to the chef to interpret the list of ingredients to realize the final dish,” said Laiskonis. “Rather than replace any one given human task, Watson merely helps to kickstart creativity and create several possible directions that the chef can then ponder and navigate in the kitchen.”