THE WORLD'S FIRST AI FLAVOUR
When we think about Artificial Intelligence and the opportunities it could unlock, we tend to picture things we can see, hear and touch. But what about taste? Does AI have a role to play in developing and honing flavours?
Geneva-based Firmenich – the world’s largest privately owned perfume and taste company – certainly think so.
In October 2020, they announced the first ever AI-created flavour: a lightly grilled beef taste for use in plant-based meat alternatives. They say it’s “delicious”. I say the proof is in the pudding (or should that be burger?!). Either way, it’s a big breakthrough made possible through Firmenich’s collaboration with Microsoft. Utilising Microsoft’s Azure Cloud Platform, it was possible to leverage raw materials from across Firmenich’s broad database.
Eric Saracchi, Firmenich’s Chief Digital & Information Officer, explained:
“The Microsoft Azure Cloud platform will enable us to harness the unique creativity of our Flavorists for enhanced formulation. We’re empowering them with a precise flavour starting point, as well as additional suggestions for optimized ingredient combinations from which they can use their rich expertise to create bespoke tastes.”
Firmenich Flavours President, Emmanuel Butstraen, added:
“The exciting addition of AI allows us to better leverage different raw materials and explore new creative leads. Taking into account specific product parameters such as 100% natural ingredients and regulatory requirements, the technology enhances our flavorists’ capability to create superior taste solutions and accelerate our product development.”
And whilst a “lightly grilled beef” taste may not sound all that exciting, the potential this innovation could unlock in the future certainly is.
For one, there’s been a huge boom in vegetarian and vegan offerings in recent years. I wrote a blog about this very topic in May, commenting:
“The big hitters in the fast-food industry are unlikely to be getting in on the act for health or ethical reasons. Nevertheless, the tide of public opinion is turning and there are big bucks on offer for those who take the change seriously.
“KFC, Subway, Burger King and many more are shaking up their menus with substitute offerings – before it’s too late. For a company literally named Kentucky Fried Chicken, that’s a really big deal. But there are big rewards to reap. According to market researchers, Mintel, the UK market for meat-free foods was worth £740m in 2018, up from £539m only three years ago. What’s more, retail sales are expected to increase to £658m by 2021.
“Companies such as Beyond Meat and Impossible Foods are leading the way with their plant-based alternatives - flavoursome and practically indistinguishable from the “real thing” once you’ve added a bit of garnish, sauce and a brioche bun... And it’s not just burgers. “Impossible Pork” was launched at CES 2020 to much fanfare. The baton has been passed from long-standing staples such as Quorn and Linda McCartney, blessing veggies and vegans with more choice and variety than ever before.”
The more plant-based alternatives taste like the “real thing”, the more likely meat-eaters are to convert. A big win for animals, the environment and human health.
And what if these AI-developed flavours could encourage children to eat healthier? Kids who’ll choose carrots and cauliflowers over chocolate are few and far between – despite the best efforts of parents around the globe. But what if “healthy” foodstuffs were tastier? Could AI flavours spell the end for fussy eaters?
McCormick & Company – an American spice, marinade and condiment producer – are another making a foray into the world of futuristic flavours. In 2019, a four-year collaboration with tech multinational IBM was announced.
Products devised the traditional way can take months – even years – to evolve from the conceptual stage to being ready for market. As a blog post by Robin Lougee of IBM explains:
“Designing new flavour experiences is an art and science requiring many years to become proficient. There are thousands of available ingredients. The product developer must not only determine which combination of ingredients to use but also the ratio of amounts needed to meet specific targets and a host of other requirements. Even the smallest change in the amount of an ingredient can make or break a flavour.
“Product developers gain expertise over years of hands-on experimentation, iteratively creating candidate formulas, manufacturing samples, running a variety of laboratory and consumer tests on the samples, and learning from the results. It is a time and resource intensive process.”
That’s where IBM’s system can help simplify the process – detecting patterns gleaned from existing formulas, results of taste tests and sales data to generate new or enhanced flavour combinations. In doing so, the speed of flavour innovation at McCormick & Company is up to three times faster, whilst delivering “highly effective, consumer-preferred formulas.”
In time, more focus will fall on developing truly novel flavour experiences. Lougee explains:
“Here, the variation or distance between ingredients (e.g. vanilla vs strawberry) is likely more significant than the choice within any one ingredient family (e.g. vanilla). Our system learns and uses a distance model to suggest desired flavour formulas.
“Because our AI system is data driven, the insights it delivers evolves as the data changes. Product developers have an overwhelming number of combinations and proportion of ingredients to choose from. They may have some go-to solutions for certain component flavour challenges. For example, because of the time sensitive nature of their work they may use their favourite “bacon” standby any time the formula calls for a bacon flavour component. Having an AI apprentice that can intelligently explore more options quickly helps them avoid using habitual standbys when evidence suggests better alternatives may exist.”
Thankfully, we can expect the flavours to be more appealing than those suggested by MIT’s “How to Generate (Almost) Anything project. According to Business Insider:
“To come up with the new pizza topping options, hundreds of recipes for pizzas from food blogs and recipe websites were entered into the AI, teaching the artificial neural network to "create" pizza combinations it "thought" could work - and this is just the beginning.”
The suggestion? A sweet potato, bean and brie pizza. Fancy something meatier? How about a pizza topped with shrimp, jelly and Italian sausage… Hmm, I think I’ll pass. But it goes to show how important thinking outside of the box is for flavour combinations. If consumer data proved – as we’d expect – that a shrimp, jelly and Italian sausage is disgusting, then the system can learn from that and tweak its output accordingly.
All our go-to favourites came about by experimentation, after all. Toad-in-the-Hole, Big Macs, Beans on Toast… Even mixing tomato ketchup with mayonnaise! Don’t like the sound of that? Don’t knock it until you’ve tried it. That one’s a game-changer…