Photograph by Matthew Ryder and Jamie Simpson
Read Caption
The ingredients for "poo- pourri" ice cream.
Photograph by Matthew Ryder and Jamie Simpson
The Plate

Software Creates New Flavors, Some Not for the Squeamish

I’m betting you’ve never opened your pantry and gone, “Huh, I wonder what would go really well with fermented cow manure right about now.”

But just in case you’re wondering, I’m told that jasmine and roses taste delicious with it.

Jonathan Brill, founder of Special Projects Agency, an innovation and design consulting firm based in Sausalito, California, is trying to shake up culinary traditions by creating a program that combines unexpected flavors based on their chemical makeups.

His nascent software, tentatively called “Why Flavors Work,” popped out the unlikely fecal/floral combination—and it’s delicious, says Brill. He quickly adds that you don’t actually eat the poop (please, please don’t eat the poop).

“You lightly smoke milk with dung from hay-fed dairy cows, and it creates this very nice light smokey hay-like flavor for the milk—and you add rose petals [and jasmine] to it. And it becomes this unbelievable ice cream,” Brill says. And just in case you’re thinking of trying this at home, Brill and his colleagues specially source and process the manure to make sure it’s safe to use as fuel.

Brill’s passion project—he calls it his “5 to 9 job”—uses data science to create a global cuisine. “Not a 1980s fusion cuisine,” he says, “but something that’s truly a mature exploration of global flavors.” He teamed up with former Special Projects Agency data scientist Jonathan Pelsis and Executive Chef Jamie Simpson of the Culinary Vegetable Institute and Chef’s Garden in Ohio to bring it to fruition.

While IBM’s Chef Watson (see My Computer, My Sous Chef) learns from a magazine’s recipes, Brill says Why Flavors Work uses a novel algorithm that considers both human physiology and food chemistry.

“‘As with most complex machine learning that’s effective, there needs to be a human feedback loop,” Brill says, to help decide which ingredients actually taste good together in the real world, and in what proportion. (The website features a few options that may be easier to wrap your mind around than the ice cream: octopus tacos with cumin, pickled jalapeños, and berry cream; salt-cured cod with cinnamon.)

But with such decontextualized flavors, you’ve got to have someone to test-drive the combinations. That’s where Simpson comes in. “Jonathan [Brill]’s the idea, the ‘what if what if what if’—and I try to use it as a program that works,” Simpson says. “We’re in the very infant stages, and if it doesn’t work, we’ll go back, and reformat and just change the way it speaks.”

A food science autodidact (his undergraduate degree is in industrial design), Brill took his curiosity about commestibles to the scientific literature to figure out which molecules compose complex flavors, and to understand the reigning theories about how best to pair them.

It’s a tall order—there’s a lot that goes into how we experience a food. There are the big five flavors we sense using receptors on our taste buds: sweet, salty, bitter, sour, umami (and maybe fatty, depending on whom you ask). The rest of taste actually comes from our noses. As we chew a mouthful of food, volatile organic compounds rise off of the masticated mess and float up through the back of our noses in a process called retronasal olfaction. This is where we get the nuance of flavor—roasted peanuts, for example, emit nearly 200 of these odor molecules.

And different theories attempt to explain why we like specific combinations of these flavor molecules. In Western cuisine, we usually like foods made of ingredients with similar molecular signatures. Think pizza: mozzarella cheese, tomato sauce, and parmesan all contain the chemical 4-methylpentanoic acid.

But in Eastern cuisine, chefs prefer to use ingredients with contrasting flavors that have very little molecular overlap (see Why Indian Food Breaks All the Food Science Rules.).

Jasmine and feces—key ingredients in Brill’s ice cream—both contain the molecule indole in different concentrations (although Brill says the algorithm is weighing so much different information that there is probably more going on than overlap of a single component in their bouquets.)

“So what we’re trying to find is not just the similarities, but the difference, and contrast,” Brill says.

“The workflow starts with an ingredient, and then it will spit out a couple ingredients and you pick one or two of those, and then it will spit out a couple more ingredients that also relate,” Simpson says. “And then your ideas, I mean the brain just starts clicking which is really cool…it’s like a fireworks show.”

These fireworks have led Simpson to create unusual dishes—like white chocolate combined with cauliflower and scallops—which Brill and Simpson are compiling into a cookbook that they are hoping to release early in 2016. They intend to roll out the software first to professional chefs and then to home cooks over the next few years.

“We’re taking thousands of years of tradition and preservation here, and serving that in this juxtaposition with something that may have never been seen in the culinary world before,” Simpson says. “It’s a beautiful balance.”