They undoubtedly have one of the best jobs in the world.
But whiskey sommeliers may soon have a competition for their work – from AI.
Scientists have made it machine learning algorithms that can tell whether a whiskey is American or Scotch and identify its strongest aroma.
And they outperform human experts, the results show.
The aroma of whiskey is determined by a complex mixture of aromatic substances, which makes it difficult to assess.
Teams of human experts are often used to identify the strongest whiskey notes but this requires significant investment in time, money and training – and collaboration between experts is often rare.
A team from the Fraunhofer Institute for Process Engineering and Packaging in Germany analyzed the molecular structure of seven American and nine Scotch whiskeys.
These were 12-year-old Auchentoshan from the Scottish Lowlands, Talisker Isle of Skye Malt 10-year-old, Jack Daniels. Tennessee Whiskey and Woodford Reserve Bourbon.
Scientists have developed machine learning techniques that can tell whether a whiskey is American or Scotch and identify its strongest aroma.
They undoubtedly have one of the best jobs in the world. But whiskey sommeliers may soon have competition for their jobs – from AI (stock photo)
To do this they used algorithms including molecular odor prediction AI they developed themselves called OWSum.
Algorithms analyzed the whiskeys based on the molecules found and identified the country of origin of each dram and its five strength notes.
The authors compared the results of the algorithms with those from a group of 11 experts.
The results showed OWSum was able to determine whether a whiskey was American or Scotch with more than 90 percent accuracy.
AI identified caramel as the dominant note of American whiskeys, and apple-like, solvent, and phenolic – often described as smoky or medicinal aromas – as the dominant notes of Scotch whiskies.
Both algorithms were able to identify the five strongest notes of a particular whiskey more accurately and consistently than any human expert.
The authors believe that their method can enable whiskeys to be classified more quickly and to identify key notes on their ethnicities.
Their findings were published in the journal Communications Chemistry.
