Scientists develop new wine-tasting AI technology

Researchers at the National Institute of Standards and Technology (NIST), based in Maryland, have created artificial intelligence able to virtually ‘taste’ wine with more than 95% accuracy.

New AI technology developed by scientists in the US has passed a wine-tasting test with flying colours, proving a “more efficient but less fun” method of sampling wines, according to its makers.

The team began by training the AI system’s virtual palate, using 148 of the wines from a dataset of 178 made from three grape varieties.

Each virtual wine had 13 characteristics to consider, including level of alcohol, colour, flavonoids, ash, alkalinity and magnesium.

Each characteristic was assigned a value between 0 and 1 for the network to consider when distinguishing one wine from the others.

Brian Hoskins, NIST physicist, said of the experiment: “It’s a virtual wine tasting, but the tasting is done by analytical equipment that is more efficient but less fun than tasting it yourself.”

The technology then had to be put to the test. Tt was given a virtual wine-tasting test on the full dataset, which included 30 wines it had never seen before.

The system passed with 95.3% accuracy, and made just two mistakes on the wines it had never seen before, which researchers saw to be a good sign.

“Getting 95.3% tells us that this is working,” said NIST physicist Jabez McClelland.

The wine-tasting AI technology has not been designed as an artificial sommelier, but to prove that similar devices could potentially be scaled up and used to build new AI systems.

These systems, using less power than previous forms, could be used to in applications such as “smart” clothing, miniature drones, or sensors that process data at the source.

An Australian e-nose prototype has previously been developed to spot the difference between cheaper and more expensive whiskies in less than four minutes. Read more on that here.