Hard to imagine a script of the movie Terminator in the Rise of the Machines started cooking burritos, but it could well start. Instead of unleashed terror among futuristic war machines self-aware, maybe all starts with a simple step: designing creative recipes. A few weeks ago, an intelligent machine Watson, launched all their cognitive capacity to invent itself innovative and sophisticated culinary products, and I knew I would delight diners, serving a multitude of burritos with black beans for chocolate soybeans, Edam cheese, meat and apricot. Beyond the initial apocalyptic joke, a machine has been able to create. No talk of proposing rational solutions to a problem, but what would be the first steps of original and creative thinking.
What makes a machine intelligent? Last Saturday, a computer program that was posing as a Ukrainian teenager named Eugene Goostman managed to fool a handful of jurors. By overcoming the so-called Turing test, announced the victory of many artificial intelligence. Mimic the responses of a child is a major challenge for a program, but advances in artificial intelligence are moving much more sophisticated and complex paths.
A cover lobster with basil, crispy pork, saffron sauce and piquillo peppers. That dish, as peculiar as delicious as those who have tried it is the creation of a computer. Not any one but the famous Watson, a supercomputer developed by IBM able to defeat irony and double meanings-something so human-to win a contest popular television. That lobster recipes with thousands more, has served to blur the boundary of what machines can not do: think creatively.
It is a food that volcarle ingredients inside to return the pudding ready to serve. Rather, it is the chef who invents suggestive dishes that nobody had ever done before thanks to their knowledge of flavors, raw materials, chemical processes and human psychology. “We’re not trying to pass the Turing test kitchen, but trying to invent new recipes”, recently summed up one of the designers of this system of computational creativity.
The manager of this team of researchers from IBM, Florian Pinel explains that landed in the kitchen looking for a field that could develop creativity Watson. “To prepare a dish there are thousands of billions or trillions of possible combinations. Watson processes them all and just picks the most interesting”, says Pinel, further systems engineer’s degree in culinary arts.
The challenge was to develop a very human creativity, ie limited to avoid providing random mixtures of flavors and ingredients and create solutions knowing that they would be surprised. The first step to that operative creativity was learning to cook, to know why we like certain flavors or what makes the quiche one quiche. “We created a database of foods containing about 30,000 wikia.com recipes, all containing the Wikipedia about food [Watson is able to understand what you read], as well as information about the ingredients at the molecular level, information and nutritional studies of taste perception in the human palate, “lists the manager of this project, which has already published several papers on the experiment, which was kept secret for a while.
Then it was to develop more complex algorithms that indicate the steps to develop creative ideas to understand what the problem to be solved, accumulate knowledge to cope, begin to suggest ideas and select the best work at all levels. With all the information that counts, Watson is able to perform all sorts of combinations for unexpected from everything you know to try. And hence arose chocolate burritos, Vietnamese kebab apple or custard pudding and smoked bacon, Pinel’s favorite: “Although what I like is not a particular recipe, but you can try a new recipe each once and never repeat the same twice, for the rest of your life”.
QUIET, YET JUST FLY
Logically, Watson gets apron or grips or pans. To implement your ideas, researchers at IBM have been working with chefs from the Institute of culinary education in New York (ICE), one of the most important cooking schools of USA. In early versions of the experiment, Watson merely suggests combinations of ingredients according to their cultural, psychological and chemical knowledge, but without indicating or amounts or further explanation in the section on the preparation of the dishes. “At first we focus on the production of new combinations of tasty ingredients, because we have worked with chefs and we know how to cook the dishes. But are adding algorithms to generate proportions of ingredients and more precise instructions”, says Pinel.
So much of the challenge was in the hands of the cooks who should interpret Watson’s suggestions. This creative collaboration between machine and human is one of the core values that IBM defends its experience: silicon neurons working shoulder to shoulder to innovate. What taste was left to the chefs had to make pinches to the computer? “There were proposals that seemed there was no way they could become a viable recipe. However, as the worked we discovered that the dish was delicious” sums up James Briscione, director of culinary development and ICE experiments Chef.
“In some cases we had to work very hard to achieve, sometimes had to perform eight or more different tests with the same set of ingredients before reaching a dish that could serve with pride. But each proposed system has been a success, we have never failed”, defends Briscione.