AUSTIN, Texas — One of the best meals I’ve ever eaten probably could not have been created without a computer. Not any average computer, mind you. Cognitive Cooking is a relatively new IBM Watson project unveiled here during SXSW that uses ingredient data and human food desires to create previously unimagined recipes — ones that real cooks then have to adjust and then make into edible food.

Cognitive Cooking is the brainchild of Florian Pinel, a senior software engineer at IBM and trained chef. The system puts computers in kitchens to complement — not replace — chefs. “It’s about a collaboration between humans and computers,” Pinel said. This computer doesn't cook; it just suggests the best ingredients.
See also: 10 Startups to Watch in 2014

IBM scientist Florian Pinel, who thought up Cognitive Cooking, is also a trained chef.
Image: Mashable / Lance Ulanoff
To achieve a delicious — and more importantly, inventive — collaboration, Watson's cognitive computer had to understand some critical characteristics of ingredients and flavors.

The scientists and chef collaborators filled the Cognitive Cooking computer with information about the taste profiles and chemical compositions of virtually every ingredient known to man. Pinel’s team also loaded the computer with recipes from public repositories like Wikipedia and the Institute of Culinary Education (ICE), which is a partner of theirs.

A key component of Watson’s cooking school education is based on research into Hedonic psychophysics, which sounds complex but is really quite simple: In various experiments, people were asked to smell ingredients and grade them based on level of pleasantness. Smell is actually more important than taste, when it comes to good food, since we only have five tastes — sweet, sour, bitter, salty and umami — but hundreds of smells. This information actually helps predict the “pleasantness” of combined ingredients. The knowledge from this research plays into Watson’s Cognitive Cooking, so it can suggest new or never-before-imagined ingredient combinations.
The act of cognitive cooking, though, is collaborative. To demonstrate, IBM invited me to Mettle Restaurant in Austin, where two expert chefs, James Briscione and Michael Laiskonis, played a kind of culinary Mad Libs. They started with a desired ingredient (like bacon) and then selected one of Watson’s many cuisines (like Hungarian). Next, they selected a dish type (gumbo). The Cognitive Cooking computer returned a list of ingredients that would work well together to create a pleasing dish. Pinel explained that the system actively seeks the most novel combinations.

That's when the hard work began. The chefs took those ingredients, which they can adjust in the system, and then figured out how to make a dish that both tastes and looks good. At dinner on Thursday, the chefs and their computer counterpart succeeded in spectacular fashion, delivering a five-course meal that included exquisite but somewhat oddly-named dishes like Czech pork belly moussaka” and duo of Italian roast duck.
IBM executives and chefs repeatedly pointed to the creative opportunities possible with this new technology. “Everything we do is based on somewhere we’ve been,” Briscione said. "It’s taking us to whole new place.”

Pinel, who launched the project two years ago with a very small team like a "pet project," thinks the possibilities for combining small components to create “novel artifacts” is unlimited. He said he could see Cognitive Computing applied to fragrance creation, personalized travel itineraries and even business processes.

Even though every single bite of my meal was five-star-worthy, I still had to ask: Did Watson’s Cognitive Cooking engine ever present them with a set of ingredients that simply wouldn't work together? Pinel and Briscione both immediately responded: midwestern mushroom stroganoff. Briscione said it tasted fine, but was definitely not a stroganoff.
Based on this example, Cognitive Cooking may not be ready for your own kitchen just yet — but based on Watson's ability to create new and delicious recipes, the future of modern cooking will most certainly be a collaboration between man and machine.
Start by telling Watson your favorite ingredient.
We like Bacon.
Then you select a cuisine.
Next, you choose the kind of dish you like or hope to make.
You get a list of ingredients that you can adjust.
Which you can then adjust before cooking.
This screen shows how, at a molecular level, the ingredients Watson has pulled together have some similarities.
There's also a mode where you can converse with Watson to achieve the same result.
More chatting with IBM Watson's Cognitive Cooking, which is always seeing to "surprise."
Have something to add to this story? Share it in the comments.