Bored AI engineers, the British Pancake Day, and lockdowns – get the picture?

British engineering AI software company Monolith used artificial intelligence to create the ultimate pancake recipe. Monolith, which is an offshoot of the U.K.’s Imperial College, found a recipe that used seven ingredients and needed just one toss for a drool-worthy stack of pancakes. (Daily Mail)

How they did it

The Monolith engineers went through over 800 recipes for American-style fluffy pancakes. They then drilled down to 31 which had enough positive and negative reviews and so could be used to train a machine learning model.

They then conducted hundreds of thousands of calculations across the three pancake-making steps of ingredients, mixture, and frying.

The firm used a machine learning algorithm called Random Forest Regression. This is a collection of decision trees working together that ultimately determine the optimal solution – in this case, the best recipe that delivered the fluffiest pancakes.

The perfect recipe

Here goes:

• Ingredients – 210 grams of flour, 48 grams of sugar, 14 grams of baking powder, six grams of salt, two eggs, 256 grams of milk, and 25 grams of butter.
• Cooking directions – follow generally accepted practice (the AI kept out of this, apparently)
• Mix the dry ingredients separately from the wet ingredients before combining them
• Do not overmix as this will make the pancakes chewy instead of fluffy
• Grease a frying pan with a moderate amount of oil or butter
• Add the pancake mix to the pan
• On an electric hob, use medium heat; gas hob, use low heat
• Flip the pancake once the underside is cooked but the top is raw
• [Monolith’s hint on when to flip – little bubbles appear on top and the edges of the pancake start to solidify]
• After the flip, the pancake should land uncooked side facing down on the pan
• Cook for another half a minute or so
• Serve!

Related Story:    Rustle Up A Nice Dinner Using AI And Plant Jammer