Google AI creates 2 yummy dessert recipes
What happens when a cake and cookie merge? Cakie. And bread and cookies make, of course, Breakie. While this sounds like random fusion food menu from a bakery or café, it’s actually not by a human, but a robot, the Google AI. We are talking about a novel experiment we came across recently on the Google blog. It was an entry by Google developer Sara Robinson, where she described how an AI/ ML experiment resulted in baking.
What was Google’s bakery project?
Inspired by the pandemic-spawned spike in searches for baking in winter 2020, the team at Google Cloud "decided to dive a little deeper into the trend and try to understand the science behind what makes cookies crunchy, cake spongy and bread fluffy," as they note on their blog.
What was the process of making the recipe?
First, these Google Cloud employees actually collected hundreds of recipes based on keyword related to bakery. They listed all common ingredients and converted recipe measurements to ounces overall for convenience. They also categorised the end results under ‘cake’, ‘cookie’, ‘bread’, putting more elaborate breads such as the ones made from banana, zucchini and pumpkin under ‘cake’. They then fed all the ingredients data into a Google Cloud too called AutoML Tables. Here the ingredients were listed in tables and analysed via Machine Learning to understand whether it should be a bread, cookie or cake.
The algorithm model they had set in the table was able to tell them about hybrid recipes, for example, it could be 50% bread, 20% cake, 30% cookie. Now Robinson and team tried baking 2 of these hybrid treats, and were surprised to find them not just edible, but yummy.
What are the implications of the project?
Overall, beyond giving some new recipes, the experiment demonstrated just how broadly the applications of AI and ML can be. What started with recipes can convert to products that all require grouping together various ‘parts’ of a project to make the ‘whole’. We would be keen to see what it leads up to.