Most importantly, the system will pluck underused ingredients or combinations—venison, bear meat, fenugreek, etc.—and add them to recipes to excite palates. Inductive reasoning will play off of human perception; the system will act as the epicuriously creative neural pathway that you wish you had. This is the same type of thinking that IBM supercomputer and Jeopardy! champ Watson used to outsmart Ken Jennings and Brad Rutter by getting the same answers, just faster and more efficiently.
IBM’s still-unnamed system is more than just a helper for stumped grocery shoppers. The creativity component also has a crucial social component. In addition to pleasing picky eaters and finding ways to substitute healthy ingredients for diet-conscious consumers, it can also help resource-poor areas make the most of their produce.
From 50 recipes of quiche, the system can infer that a “good” combination of ingredients for any variation of quiche would include eggs, at least one vegetable, and three spices. To generate these food leads, if you will, AI cross references three databases of information:
A recipe index containing tens of thousands of existing dishes that allows the system to infer basics like “what makes a quiche a quiche”
Hedonic psychophysics, which is essentially a quantification of whether people like certain flavor compounds at the molecular level
Chemoinformatics, which sort of marries these two other databases, as it connects molecular flavor compounds to actual foods they’re in
IBM isn’t trying to create a robochef to rule us all. Rather, the system is supposed to a handy companion to curious cooks. To demonstrate this, the Institute of Culinary Education provides professional chefs to test the computed recipes. The result is relatively successful, Varshney told Fast Co.
“Nothing is really crazily bad, though there are certainly things we’ve tried making that weren’t spectacular, like a mideastern mushroom stroganoff,” he said. “Out of the 20 most recent dishes, 17 or 18 have been really good.”