鶹ý

To recreate delicious meals, don’t treat recipes simply as algorithms

Recreating the delicious meals of my youth involves balancing the algorithms of online recipes against my recollection of watching my father cook, says Annalee Newitz

IT’S the time of year when my circadian clock starts yelling at me to make cookies and latkes. Unfortunately, I don’t know any recipes for them. I blame my father, an incredible cook who bought every new kitchen gadget he could: fancy mixers, strange attachments for his food processor, ice cream makers, pasta squishers and even an espresso machine that he plumbed himself so that it would automatically refill with water. The problem is that growing up with a parent like that means delicious meals appear by magic every night. You have very little incentive to learn cooking.

My dad didn’t pass along any family recipes – and, honestly, I’m not even sure we had any. Although my grandmother loved to cook too, she didn’t bequeath the secret of latkes to us. My dad bought a book of Jewish recipes to figure it out, which is perhaps the quintessential second-generation immigrant thing to do in the US. His parents worked so hard to assimilate that they didn’t think it was important to teach my dad Yiddish or how to make latkes, knishes, kasha varnishkes, matzo ball soup and all of the other European Jewish dishes that I yearn for in winter.

“I weigh the algorithms of online recipes against my recollection of watching my father produce a plump, circular pancake”

I didn’t start cooking in a serious way until after my father was dead, so I will never know if there was some secret family recipe he never revealed. But there’s still one lesson I hold dear from years of watching him cook. A recipe is an algorithm – perhaps one of humanity’s oldest consciously produced algorithms – and yet cooking also transcends the power of algorithmic reason.

These days, when I want to cook anything, I start with recipes online. This is the algorithmic part of my process, where I hunt around for variations on the instruction set and parameters. One person likes to add flour to their latkes for structural integrity; another includes matzo meal; yet another insists that it’s all about the egg. I don’t know exactly which elements will produce my father’s results, but I know generally how things should look and smell when a recipe is going the way I like. I weigh the algorithms against my recollection of watching my father produce a plump, circular pancake as opposed to one with a ragged, crispy edge.

When engineers write algorithms, they do something very similar. They will compare the output of an algorithm to output from a real-world process, asking whether, for example, a produces clouds that look natural. Still, cooking requires me to compare recipes to things that are ineffable: a smell, a taste, a memory. Argue with me if you want by saying that we can stick smells into olfactometers and . That may be true, but the taste of food is profoundly subjective.

If I had a digital file with the flavour of my father’s latkes, it’s possible that the resulting food wouldn’t taste right. Because what I’m trying to make aren’t perfect reproductions of a set of chemicals. They’re my memories of a small, warm kitchen where onion and potato sizzle in chicken fat, the bite of black pepper on my tongue and a crunch that cannot be quantified.

Topics: algorithms / Food science