IBM is finding an answer to a tough question of its own: Can it give computers a sense of taste?
Four years ago, researchers at IBM set out to answer this question, starting first with a liquid that is largely free of taste and can be difficult for a human to tell apart: water. After the computer became familiar with different water samples that had slight variations in composition due to their mineral content, it was put up against human competitors to see who could better identify a water sample it had seen before. The computer won.
“The artificial intelligence system was better than human tasters at distinguishing specifically four different kinds of mineral water,” said a researcher on IBM’s artificial intelligence-assisted e-tongue technology called Hypertaste.
The circular-shaped disc undergoes its “taste test” by creating a digitised chemical composition of a liquid it has sampled. The “fingerprint” is then compared with other liquids in a database using artificial intelligence — a process that takes less than a minute to identify a match.
While Hypertaste is still a few years away from widespread commercial use, IBM is working with industry partners on different applications for it. These include collaborating with food and beverage companies to capture and predict different kinds of flavours, and quickly identifying coffee, soft drinks, and other offerings that would resonate with consumers.
Hypertaste is not meant to replace human experts, but rather to offload some of the most mundane or difficult tasks, like repeated taste testing of a product, to ensure quality remains the same from batch to batch. It can also be used to assist in product innovation by discovering new and different flavours.
Future applications for the technology could include finding contaminants in drinking water, tracing raw materials throughout the supply chain, and checking for the presence of foodborne illnesses.
The more people that are using it and can benefit from it and feed the data back into it, the better the technology will get.
Artificial intelligence is an increasingly popular tool for food companies. With countless bits of information and pressure to innovate and get products to market faster, companies are no longer just relying on traditional methods of R&D and testing.
McCormick & Co. partnered with IBM three years ago to comb through data faster and more effectively by knowing which ingredients work together or which ones can be used as substitutes for each other. McCormick has introduced eight products so far, all created with the help of artificial intelligence.
It allows “our product developers to explore flavour territories more quickly and efficiently to learn and predict new flavour combinations across sensory science, consumer preference, and flavour palettes,” the company said in an email.
Conagra Brands also uses artificial intelligence-enabled platforms to identify consumer preferences and release on-trend products to the market in a fraction of the time. Dairy giant Danone has announced a collaboration in 2020 to use artificial intelligence to unlock hidden nutrients in soybeans.
A ‘major‘ competitive advantage
The average human tongue has up to 8,000 taste buds that help distinguish the five tastes of sweet, sour, salty, bitter, and umami. However, the human nose has a strong influence on how people actually taste food. Not surprisingly, the two senses are closely connected. For example, the taste buds on the tongue determine if a food is sweet, while the nose determines what that sweet is connected to, like a strawberry or a grape.
Biotechnology company Aromyx estimated that while the average human nose can distinguish a trillion different odours, it often has difficulty deciphering the difference between these smells and tastes.
To solve this problem, Aromyx has turned to artificial intelligence. The company has cloned receptors found in the nose and tongue. Scientists then place something a consumer might taste, like a drop of coffee, on those cloned receptors before measuring how much they turn on—in effect, replicating the reaction that occurs in the brain when a person takes a sip of his or her morning brew.
The process, done millions of times over, allows Aromyx to assemble a database for an outside R & D team to use when testing coffee. Based on how those receptors react in the future, a CPG could determine if consumers will enjoy a similar kind of coffee; how their reaction might differ based on things like demographics or even where they live; and why the person does or does not like it — say, because it has a floral flavour or a certain ingredient.