Maltesers, Marmite and ML – Google gets baking with help from machines

Share on facebook
Share on twitter
Share on linkedin
Share on email

When companies collaborate it’s often a collaboration that makes sense. For example, Samsung and AKG is a pairing that makes sense, one makes smartphones the other tunes earphones for those smartphones.

But today’s collaboration is, there’s no nice way of putting this, it’s odd.

Google has partnered with Mars and Sara Robinson, the winner of the Great British Bake Off to create one of the strangest sounding recipes we’ve ever heard of – Malteser Cakes with Marmite butter cream icing.

As somebody who loathes Marmite, this sounds like the thing of nightmares. So how did this nightmare become reality?

While GBBO is Robinson’s claim to fame, she is also a machine learning engineer.

So when the pandemic hit last year Robinson did what any geeky baker does – you mix tech with confectionary.

The engineer trained a machine learning model to generate recipes for cookies, cakes, scones, traybakes and combinations of each.

Meanwhile, Google looked at top search trends and discovered that a popular question was “Is Marmite sweet or savoury?”. The answer, as we all know, is that Marmite is wrong.

Okay we’ll stop with the Marmite gag.

In order to find out the answer to that age-old question, Google decided to combine Robinson’s model with Cloud AI Platform Notebooks to engineer features and further develop the model to find a sweet recipe containing Marmite and Maltesers.

“Our goal for the project was to build a model that could provide the foundation for us to create a new recipe featuring Maltesers and Marmite. To develop a model that could produce a recipe, Sara wondered: what if the model took a type of baked good as input, and produced the amounts of the different ingredients needed to bake it,” wrote Google.

“Since Maltesers are primarily sold in the UK, we wanted the recipe to use ingredients common to British baking, like self-raising flour, caster sugar, and golden syrup. To account for this, Sara used a dataset of British recipes to create the model. The dataset consisted of four categories of popular British baked goods: biscuits (that’s cookies if you’re reading this in the US), cakes, scones, and traybakes.”

Robinson’s model was further refined using Google’s AI Platform Hyperparameter Tuning and finally deployed using AI Platform Prediction.

From here the model was able to formulate recipes but adding the Maltesers and Marmite was a decision made by Google and Robinson alone. The final recipe combines a cake and biscuit recipe and the whole affair is topped with Marmite butter cream and crushed Maltesers.

What makes this whole affair so impressive is that AI was able to generate two competent recipes that work. We can’t speak to what the recipe tastes like but given our experience in the kitchen, this looks like a rather tasty cake and biscuit base.

That’s impressive because baking is really a science meeting art. Balancing ingredients in such a way as to get the cake to rise and have a good crumb structure aren’t easy things to do and the fact that a machine learning model can produce a unique dessert is very interesting.

You can find the recipe for Maltesers AI Cakes (4d61727360a) in PDF format here.

Maybe we’ll give this recipe a try over the coming long weekend, Marmite and all.

[Source – Google]

Brendyn Lotz

Brendyn Lotz

Brendyn Lotz writes news, reviews, and opinion pieces for Hypertext. His interests include SMEs, innovation on the African continent, cybersecurity, blockchain, games, geek culture and YouTube.