AI put to new, novel purpose — helping spice makers develop new flavors

When considering the current and possible applications of artificial intelligence, there are a host of industries that leap to mind as obvious candidates for improvement or upheaval. Medicine, for one. Image recognition and editing, another. We’ve chronicled some other, less obvious industries too, but they at least make sense given how technical their purview happens to be. But every once in a while, we hear or read about a new application that truly surprises us (and is a testament to the wealth of innovation happening in this field every day). To that end, one of the newer applications that caught us delightfully off guard? The world’s largest spice maker using AI to help it develop new flavors.

How can a machine help with flavors?

So with something like image recognition or medicine, it can be pretty obvious how A.I. will help practitioners of those fields. You have boatloads of historical data to feed neural networks and a quantifiable and easily determined ‘answer’ both humans and the A.I. itself can test its outputs against. That doesn’t mean doing any of these thing successfully is a walk in the park (by no means is it, to be clear), but it’s at least easy to see how and why A.I. would be a strong candidate to help in these industries. But something as inherently human as taste? That seems to be something else entirely.

According to Axios, the world’s largest spice maker is doing just that, though: “McCormick… has begun working with IBM Research to create new spices that humans might not consider. Among its latest concoctions — the cumin pizza, says Richard Goodwin, principal research scientist at IBM.”

As it turns out, A.I. can be programmed without the same biases human chefs may have (that’s not to say A.I. is without bias — it’s not — but you can take that into account and actively attempt to counteract those tendencies). And in a field that rewards creativity, throwing a bunch of combinations at a wall to predict what might stick can absolutely yield fascinating, novel results.

It’s also worth noting that there are similarities and, dare I say, formulaic ways to predict success of recipes — mixing salty flavors and a crunchy texture, for instance; or blending savory and sweet. And in the same way that Pandora broke down every song to its base, musical elements to recognize patterns in the types of songs you’re likely to respond to, so too can computers and A.I. break down flavor profiles and combinations in search of mathematically creative but objectively tasty recipes and flavors.

To wit: “The system, which is still in the testing phase, pulls from decades’ worth of data on spices to identify a base formula for a flavor category (such as a BBQ sauce). Then it incorporates new, sometimes surprising ingredients, as well as sales and trend forecasts, to make sure the new flavors perform well.”

Why does this matter? Well, CNN reports that the algorithm can cut down on spice development time… to the tune of two-thirds.

That’s a real deal business case.