Guests at a dinner hosted by IBM Research Thursday night at SXSW were treated to, among other dishes, this Italian roast duck. Each of the six dishes was designed, in large part, by IBM’s Watson cognitive computer.
AUSTIN, Texas — We know that IBM’s Watson computer is the world’s best “Jeopardy” player. But does it have a clue about what tastes good?
On Thursday night, I was one of a small and very lucky group of people invited to an IBM event to find out.
Called “Cognitive cooking,” the event was a demonstration of Watson’s so-called “computational creativity.” Essentially, the idea went, IBM set out to have its famous computer help design a gourmet meal, one filled with dishes made from recipes the world had never seen before. And, yes, it needed to be tasty.
But let’s step back two years.
Fresh off its “Jeopardy” glory, IBM Research set out to broaden Watson’s appeal. The first, and most important, direction was health care. But Big Blue also wanted to show that it could use Watson to partner with people to create all-new things, not just analyze and interpret things that were already known, said Steve Andrews, director of the Watson group.
Another goal was finding ways to showcase Watson for the general public. In other words, though winning at “Jeopardy” was impressive and fairly easy to understand, IBM wanted to show how the computer system could be applied to every day problems. “Some people say that the crown jewel of human achievement is creativity,” Andrews said. “So we wanted to show people that cognitive computing could be taken beyond [just] answering questions.”
The aha moment was to focus this particular computing energy on cooking.
In 2012, IBM approached New York’s Institute for Culinary Education and asked if they would be interested in working together. The project? To come up with recipes that had never existed before, Andrews explained, “created in cooperation between the cognitive cooking system and [ICE] chefs.”
One can imagine the ICE people evaluating the proposal and responding the way most of Thursday night’s guests did when we got our dinner invites: “Um, yes. Please.”
This was no easy task. The first thing was for ICE to let IBM and Watson have access to its 30,000-plus item recipe database. “It was like reading a giant cookbook,” Andrews said. The idea here was for Watson to begin to understand the nature of different cuisines to be able to distinguish between, say, Spanish food and Chinese food. And to be able to recognize the kinds of ingredients that are commonly used together in certain kinds of dishes. More important, maybe, was for Watson to understand what made a particular dish different, Andrews explained, than, say, soup.
Over time, Watson learned a “tremendous” amount about food science and food chemistry at the molecular level, all the while compiling a new culinary competency that would help it develop all-new recipes based on patterns it learned. But also throw in a surprise that would leave diners smiling. At least, that was the goal.
Lest you think that someone connected Watson to a giant stove and let it make dinner, that’s not quite what was meant to happen. Rather, the IBM people and the ICE people worked on giving Watson a set of inputs — three main criteria: a dish’s geographic region; its type (a stew, or a soup, or a salad, for example); and the main ingredient. Once it had those in hand, Watson would think and think and then come up with the ingredients list that the chefs would use. The humans in the picture would be free to decide on quantities of each ingredient to use. “It’s really a partnership to create a culinary experience,” Andrews said.
So, you’re probably wondering: Does it work?
After having eaten every single morsel provided to me at dinner — a total of six courses — I can say quite clearly that the answer is yes.
Would this food scare Thomas Keller of French Laundry and Per Se fame? I doubt it. But there are a lot of nice restaurants out there that would be happy to serve the food we ate Thursday night.
It started with a Czech pork belly moussaka, moved on to Kenyan Brussels sprouts, then a Russian beet salad, two different takes on Italian roast duck (using the same ingredients for each), and finally an Ecuadorian strawberry dessert.
Each dish was tasty, though the flavors for many of them lacked a bit of punch. I’m no foodie, and certainly no food reviewer, so I’ll leave that job to others, but I will say that the presentation was lovely, the smells were terrific, and the taste was very nice. The dessert was amazing, with several of us agreeing it was the star of the evening.
During the meal, ICE chefs James Briscione and Michael Laiskonis explained the process behind each dish. They talked about the surprise of mixing particular ingredients — for example, adding avocado oil to the strawberry dessert, and what was clearly the thrill of seeing what came out of the kitchen. “Brussels sprouts with cardamom, there’s something you can take home with you,” Briscione said. “Certainly, there are elements that [I'll be taking] away with me.”
At South by Southwest this week, attendees will be able to take something away from the project with them as well. All week, IBM will have a food truck set up near the Austin Convention Center, and is inviting people to come by and experiment with the system. They can suggest cuisine, and people can vote on the suggestions. Each day, IBM will tally the votes and come up with a single dish for the next day and Watson will produce the set of ingredients for it. You can see the project in action at IBMFoodTruck.com.
But back at dinner, what did others think of the experience? Did they like it? Were they scared of top cuisine designed by a computer? Judging by the non-stop eating, and the frequent utterings of “oh, good,” it’s safe to say IBM is on the right track.
“It was very interesting, and certainly connected with things I already think about in terms of our humanity and how we maintain that in the face of technology obsession without denying upgrades and advantages that technology can give us,” said comedian and author Baratunde Thurston. “I love the idea of using robots and Skynet to help us be more creative or to be creative in different ways….Tonight, for me, was a good example of how we might expand the way we see the world and not replace what we see as human in the world.”
Also, Thurston added, “It was just delicious.”
Article source: http://feedproxy.google.com/~r/cnet/pRza/~3/H4UEWGfv5Lo/