Baking is as much a science as it is an art. Maybe to find out if the former is more important, Google Cloud AI is taking on one Great British Bake Off wins in a dessert face-off. Sara Robinson, an amateur baker and spokeswoman for Google Cloud developers, built a machine learning model that researched hundreds of baking recipes (including those for traybakes, cookies and scones) to help her come up with a new one.
The model generated lists of ingredients and quantities that were used as a basis for step-by-step recipes. The model was able to come up with hybrid recipes, and Robinson chose one that had a machine-learning cake batter on top of a machine-learning cookie.
She experimented with the model and baked to make a desert that contains Maltesers, a malt chocolate sugar similar to Whoppers. The result was Maltese̵
Both Robinson and Sawkins used some unusual ingredients to balance the sweetness of the chocolate with salt and umami. Robinson added a buttercream frosting from Marmite after Google Trends identified that “Is Marmite sweet or salty” was a top search related to “sweet and salty.” Sawkins, on the other hand, added soy sauce to a recipe he adapted from one of his mother’s creations. You can download both recipes and bake them at home to find out which one is your favorite, or get AI to discover a recipe for you on this promotional site.
Sure, AI systems are incredibly useful for things like tracking wildlife and guiding visually impaired people. But in order to actually generate better dessert recipes than the ones we mortals can come up with alone, this model must learn to leave out Marmite. Yuck.