Friends, I’m not typically a picky eater. But there’s something about Google’s new AI-generated cake recipe that... makes my skin crawl just a bit.
According to Google, the recipe came about after the company was approached by Mars Inc, the brand that you probably know as the face behind Milky Ways and Twix candy bars. In this case, Mars wanted Google to whip up something using its Maltesers candy brand, which is kind of like the UK version of the Whopper brand here in the U.S. Spearheading the Maltesers mission was Google Cloud developer Sara Robinson, who had previously spun-up machine learning models that spat out recipes for different chimera-esque baked goods like the “breakie” (half brownie, half cookie) and the “cakie,” (half cake, half cookie).
Because Maltesers are a quintessentially UK snack, Robinson wanted to create her latest ML model from the recipes of four different (but equally British) pastries: biscuits, cakes, scones, and traybakes—aka brownies, blondies, or any other baked snack that can be made in a tray. After looking at the recipes the resulting machine learning model spit out, the Google team decided that sandwiching a few Maltesers between some ML-generated cake batter and an ML generated biscuit was the way to go
All of this sounds great, but then... the Marmite comes in. For those unaware, Marmite is a dark, salty spread made from concentrated yeast that’s either Perfectly Fine or something that tastes not unlike downing an entire bottle of soy sauce, only to gag some of it back up. In other words, it’s divisive. It’s also extremely (extremely!) British, which means, naturally, Google needed to incorporate it somewhere in the recipe—in this case, as part of the frosting.
Personally, I fall on the entire-bottle-of-soy-sauce part of the spectrum, which is why I need to ask: Why ruin a delicious whopper-y cake with what’s basically solid salt in jam form? Why not go for some other form of English fare, even if 99% of them are mostly made out of meat? I know I’m likely exaggerating the salt-factor here—the actual recipe calls for little more than a teaspoon of the goopy stuff—but that’s honestly a teaspoon too many for me.
If you have a stronger stomach than me, you can check out Google’s recipe for yourself here.