Here’s the background. For about two years, IBM has been working on a way to harness Watson’s data-driven computing into more creative fields–the kinds of things where, unlike a game show, there’s no one right answer. The first experiment with that has been in the kitchen. By mining a database of freely available online recipes (as well as recipes from professional chefs and a molecular textbook) and estimating which ingredients might combine for a dish pleasing to a human palate, Watson has been creating unlikely culinary works. The quintillion possibilites–seriously, quintillion–are narrowed down and ranked by presumed tastiness and novelty.

With that data uploaded and organized by type of food, regional origin, and tastiness, the company designed an app that can make logical decisions on what might make for a good dish.

A person piloting the app starts with an ingredient; I chose bacon during a demo from IBM Watson Group researcher Patrick Wagstrom. (Because I am in Austi, I have been walking, and I am hungering for grease.) After that, I selected a region, opting for something English with influences from another country. Watson spit out a list of potential dishes it could make with those restrictions–it seems to be some kind of Michelin-star-worthy soup auteur–and I went with a quiche. Following a second of number-crunching, it showed me a list of ingredients it was planning on using. (Wagstrom admits it’s bugged out a few times in this section, forgetting dough, etc.) A few tweaks later, I had a recipe for a respectable-sounding quiche made with comte cheese.