German researchers developed two machine-learning algorithms that can determine the country of origin for different whiskeys and analyze their aromas. Whiskey aroma is a complex mix of fragrant compounds, which makes evaluation challenging.
The study found the AI even outperformed human experts in accuracy.
What Did the Researchers Report?
Having human sommeliers evaluate whiskey usually takes a lot of time and money, and experts rarely reach consensus.
To spare the whiskey industry the time, expense, and arguments, a team from the Fraunhofer Institute for Process Engineering and Packaging in Germany developed an AI system for expert whiskey evaluation called OWSum.
During testing, the system was tasked with determining the molecular composition of seven types of American whiskey and nine types of Scotch whiskey.
The AI analyzed samples and identified each whiskey’s country of origin along with its five strongest flavor notes.
The authors compared the AI’s results with a control group of 11 experts, according to the Daily Mail. OWSum determined whether a whiskey was American or Scotch with over 90 percent accuracy.
The AI flagged caramel as the most characteristic note of American whiskey, while apple and phenolic notes (often described as smoky or medicinal) were the most distinctive for Scotch whiskey.
Both algorithms pinpointed the five strongest notes of a given whiskey more accurately and confidently than any expert in the control group.
The results were published in the journal Communications Chemistry.
