Hold the glass by the stem and gently swirl the whisky. Bring it to your nose and sniff, mouth slightly open. Maybe add a little water. Now take a sip and move the whisky around the tongue. Swallow slowly….

Or just ask AI.

Satnam Singh, an Indian research scholar in Germany, and his collaborators have combined chemistry and computing to create an artificial intelligence tool that can tell whether a whisky is of American or Scotch origin with accuracy greater than 90 per cent.

And it outperforms human experts at assessing a whisky sample’s strongest aromas, correctly predicting whether it’s apple-like, caramel, fruity, solvent-like, pear-like, roasty, vanilla....

Professional whisky tasters and other connoisseurs of the “water of life” needn’t worry about their jobs or bragging rights just yet.

“We see our methods as a way to enhance efficiency in human sensory assessments — not yet as a means to end all use for the human nose,” Andreas Grasskamp, a scientist at the Fraunhofer Institute for Process Engineering and Packaging who supervised the research, told The Telegraph.

A unique mix of, on average, more than 40 compounds imparts a whisky its characteristic aroma, based entirely on its molecular composition.

The whisky industry uses panels of human experts to identify the strongest notes of a whisky. But training humans for the job means time and money. And even then, members of a panel might provide different assessments on the same whisky.

Relying on the human nose to assess flavour is subjective, Grasskamp and his colleagues underscore. The other senses, individual experiences, personality, and biological circumstances can influence a taster’s perception.

The Fraunhofer team assessed the molecular composition of seven American and nine Scotch whiskies using two so-called machine-learning algorithms, or self-learning pattern-finding software, training the algorithms to identify each whiskey’s origin and five strongest notes.

Singh worked with team member Doris Schicker to develop and test the AI algorithms on the whisky samples.

The scientists then compared the results from the algorithms with those from a panel of 11 expert whisky tasters.

“The human nose provided the ground truth and was indispensable for our results,” Grasskamp said. “We trained our models with the human nose as the reference. Still, we think the consistency with which the algorithms work as opposed to the subjective nature of human senses provides a great opportunity for quick assessments.”

The whiskies used in the study included Auchentoshan, Bowmore, Glengoyne, Talisker Isle of Skye Malt, Laphroaig, Glenfarclas, Glenkinchie, Glenmorangie Original, and Johnnie Walker Red Label from Scotland. The American whiskies were Bulleit Rye Frontier Whiskey, Four Roses Single Barrel, Jack Daniels Tennessee Whiskey, Wild Turkey 101 Proof, Woodford Reserve Bourbon, Knob Creek and Maker’s Mark Whiskey.

Multiple research groups have in the past tried to automate aroma and odour classification. In Calcutta, electronics engineers at Jadavpur University collaborating with the Tea Research Association in Jorhat, Assam, had in 2012 developed an “electronic tongue” to evaluate the quality of tea.

“This is not the first time a classification approach has been used to determine the origin of whisky, but this is the first time an algorithm could outperform human
sensory assessment,” said Singh, lead author of the new study, published on Thursday in the journal Communications Chemistry.