An (Artificially) Acquired Taste
Applying Machine Learning to discover, evolve and sustain flavours in the food and beverage industry
Ketchup flavoured ice cream exists? Really?
Yes. In fact, an Irish gelato created this flavour as a tribute to Ed Shereen. Some claim it tastes like a frozen Bloody Mary with a drizzle of ketchup on top. That’s not all, sushi flavoured KitKat and Wasabi Doritos are a thing too, in case you weren’t aware. In a world of experimental and evolving flavours, we can only begin to imagine the potential of artificial intelligence when it comes to taste.
If you frequent Starbucks you probably find yourself trying a new concoction, passed around cute tiny glasses, every few days. Imagine sharing your opinion about it with a bot over chat messenger. But instead, you’re talking about a beer you tasted at a bar. And the next time you visit and order a pint, your feedback has already made its way into a fresh batch of that brew. Intrigued?
A group of machine learning specialists and consultants in London have developed a brew label called IntelligentX — “the world’s first beer brewed by artificial intelligence to show the world what can be achieved with data.” They distribute a series of beers across stores and bars, simply classified as Golden, Amber, Black, and Pale. Each new batch is fine-tuned by tweaking ingredients, recipes or processes, based on feedback learnt by their AI technology. They use data to improve the product experience with every iteration. They’re not aiming to achieve a perfect brew, they say, but to take consumers on a journey by pushing their taste-boundaries and showing them what craft beer is capable of.
That’s about evolving flavours, what about creating it? It all comes down to the science behind flavour creation which is based on the chemical composition in ingredients. Unique combinations of such chemical compounds give food their characteristic flavours. Researchers and food scientists believe that ingredients gel together well if they have enough flavour compounds in common. Moreover, cultural recipes show that this is particular true for the West. While Eastern recipes combine ingredients with less overlapping compounds.
What does all this mean for the future of flavours?
IBMs Watson uses this concept alongside immense data of recipes, ingredient chemical compositions and established pairings to create new recipes by repeatedly working with unusual combinations and learning what works and what doesn’t. In fact, they’ve pubished a cookbook of such AI generated recipes which explores perplexing pairings like the Vietnamese apple kebab, with the vaunted mushroom-and-strawberry and the shrimp cocktail, which is a beverage with actual shrimp in it.
Palatability isn’t just confined to taste. Aroma plays an important role in defining flavour as well. Foodpairing operates on the science that only 20% of flavour we experience is derived from taste or touch. Smell is far more important. This company factors in aroma when applying data analysis and machine learning to pair ingredients and discover new recipes. While they seem to focus their services towards professionals in the food industry, a Copenhagen based AI startup PlantJammmer uses this concept to suggest new vegetarian recipes to home cooks.
At other times, the need of the hour is to sustain taste and flavours.
Breweries, for instance, maintain quality and consistency by analysing large amount of data pertaining to various brewing conditions. AI powered technology Gastrograph, created by Analytical Flavor Systems Inc., digitises this process to prevent any kind of ‘flavour drift’ from occurring. The technology aims to catch process flaws and deviations, even digression in taste or aroma that perpetrate through time, in real time and fix them before the batch is ready. This way brands can set up breweries in different geographical regions and still maintain control over the outcome, providing consumers a consistent experience.
Researchers at MIT Media Lab Open Agriculture Initiative (OpenAg) are taking climate, rather micro-climate, control a step further. Tech-greenhouses called ‘food computers’ create a controlled environment to grow food indoors. Fitted with sensors, actuators and machine vision, these greenhouses use machine language to analyse a million data points collected over each growth cycle to optimise and create ‘climate recipes’ which control detailed elements like light intensity, salinity in water and nutrients to be added.
So instead of importing dragon fruits from Vietnam, a local greenhouse can simulate the exact environment needed to grow them; even enhance it with each yield.
Global food and beverage companies invest a lot in sustaining the flavour of their branding across regions. But that’s not all, they also need to ‘glocalise’ and tailor the taste to fit different local cultures. Wonder why Coke tastes a bit different in different countries? Local water and source of sugar contribute to the level of sweetness, but so does cultural preference. Asian countries prefer their coke a bit sweeter compared to the West.
Cultural impact on food preferences determine ‘glocalisation’ strategies when brands venture into new markets. Enter, data analytics.
With every second of our daily lives being digitally documented online there’s no dearth of data sources to study global culture. Take the popular and semi-additive social platform, Instagram, for instance. Many of us are guilty of religiously clicking and posting our food on social media before every meal. Photoworld.com studied over 100,000 such photographs along with their ‘hashtags’ to discover the worlds favourite foods. According to their research, Londoners love burgers and New York salivates over bacon.
Let’s add data-mining to the mix. And not stop at analysing only hashtags.
A project at Cornell University in Ithaca, New York, applies machine intelligence to data-mining over 100 million Instagram photos to peek into cultural, social, and economic factors that shape clothing choices around the world. Facial recognition filters out images to analyse and a machine-learning algorithm identifies clothing elements and accessories. Consider applying a similar concept to food and beverages, to study beyond just hashtags? Perhaps using natural-language processing to analyse the sentiment in the captions describing these images. What kinds of new questions about global food culture can be answered?
What about us? You and I. The foodies and tipplers.
All this leaves us with a myriad of choices that our mind struggles to cope with. Bangalore based Dishq comes to the rescue with AI-powered recommendation and personalisation technology for food services. They work with a database of over 100,000 dishes, ingredients, cuisine styles, etc. along with consumer behaviour analytics, global food research and machine learning algorithms to develop APIs for food service businesses, like restaurants and delivery platforms.
What if a personalised suggestion machine like this were to evolve into a universal taste passport? Where the magic of data and AI would fuel every meal. Guiding you to a bar or restaurant that perfectly fits your taste. Ordering food that compliments your current mood. Perhaps, taking into consideration the weather around you or even the company you’re with. Less garlic please, you’re on a date! Oh, of course the chef would have all the information about your allergies and taste profile. And the bartender would mix a new cocktail put together by AI tailored to your liking, that pairs beautifully with the rest of your meal.
Doesn’t seem like far fetched dream to me.
Thank you for reading!
If you found this post interesting, please clap away — it will be extremely motivating.
This article is part of a series co-conceptualised with my friend, Symran Bhue. Every once in a while we choose a topic to discuss, debate and dive into but with different perspectives. She puts on her marketing hat, while I get tech-curious about artificial intelligence.
References:
This story is published in The Startup, Medium’s largest entrepreneurship publication followed by 336,210+ people.
Subscribe to receive our top stories here.
