"Behold the cakie: It has the crispiness of a cookie and the, well, 'cakiness' of a cake."

So says a triumphant blog post by Google Cloud's developer advocate and an applied AI Engineer for Google's Cloud AI. "We also made breakies, which were more like fluffy cookies, almost the consistency of a muffin" (or bread).

Food and Wine explains the project (in an article shared by Slashdot reader John Trumpian): Inspired by the pandemic-spawned spike in searches for baking, 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," according to a post on their blog. Then, once armed with that machine learning knowledge, they attempted to mix these attributes into what they bill as "two completely new baking recipes...."

[T]hese Google Cloud employees organized about 700 recipes covering cookies, cakes, and breads — standardizing measurements, isolating the key ingredients, and re-categorizing things like banana breads that aren't really "breads." Then, they fed them into a tool called "AutoML Tables" to create a machine learning model that was able to predict whether a recipe was a cookie, cake, or bread based on its ingredient amounts. ["If you've never tried AutoML Tables, it's a code-free way to build models from the type of data you'd find in a spreadsheet like numbers and categories — no data science background required," explains the blog post.]

Of course, recipes don't necessarily fit perfectly into one category. As Sara Robinson, who led the project, explained, a recipe might come back as 97 percent bread, 2 percent cake, and 1 percent cookie. So what if she asked the model to create its own recipe: something that's 50 percent cookie and 50 percent cake?

That's how the Cakie was born. And she was happily surprised by the results. "It is yummy," Robinson said. "And it strangely tastes like what I'd imagine would happen if I told a machine to make a cake cookie hybrid." Based on that success, she and colleague Dale Markowitz continued to tweak their model — which resulted in the Breakie.
"We should caveat that while our model gave us ingredients, it didn't spit out any baking directions, so we had to improvise those ourselves," the blog post explains. "And, we added chocolate chips and cinnamon for good measure." Robinson also built a prediction-making web app to help quickly experiment with different ingredient ratios.

They ultimately identified which ingredients were the biggest "signal" of cake-ness, cookie-ness, and bread-ness, concluding that "In our case, the amount of butter, sugar, yeast and egg in a recipe all seemed to be important indicators..."