AI can distinguish between Whisky scents with 90% accuracy new study shows

Just when you thought the AI world couldn't get any crazier, researchers have developed an AI system capable of predicting the nuanced scents of whisky with impressive accuracy.

Led by Dr. Andreas Grasskamp at the Fraunhofer Institute for Process Engineering and Packaging in Germany, the study highlights AI’s potential to complement human expertise. “The beautiful thing about the AI is that it is very consistent,” said Grasskamp. “You have this subjectivity still in trained experts. We are not replacing the human nose with this, but we are really supporting it through efficiency and consistency.”

Deciphering a whisky's aroma is a complex process. The prominent scents are typically a complex mixture of chemicals that interact within the nose, sometimes masking each other to create a unique aromatic impression. This makes it extremely challenging to predict a whisky's scent based on its chemical signature alone.

The experiment

The team analyzed the chemical makeup of 16 whiskies from the U.S. and Scotland, including brands like Jack Daniel’s, Maker’s Mark, Laphroaig, and Talisker. Using aroma profiles developed by an expert panel of 11 tasters, researchers trained AI algorithms to predict the five dominant aromas and identify each whisky’s origin.

The AI’s results were striking. One algorithm achieved more than 90% accuracy in distinguishing between U.S. and Scottish whiskies. On average, the AI consistently identified the five strongest notes in each whisky more accurately than any individual member of the expert panel. The findings were published in Communications Chemistry.

Key compounds like menthol and citronellol helped the AI identify U.S. whiskies, often associated with caramel-like aromas, while methyl decanoate and heptanoic acid pointed to Scotch whiskies, known for their smoky or medicinal notes.

Broader applications

Beyond whisky, the technology could have applications in detecting counterfeit products, ensuring consistency in beverage production, and even managing odors in recycled plastics. “The flavor notes of a whisky brand could be quickly checked from batch to batch or blend to blend based on the chemical signature alone, to try to ensure a consistent house style,” said Dr. William Peveler, a senior lecturer in chemistry at the University of Glasgow.

Challenges

Despite its promise, the AI system faces limitations. The study’s small sample size leaves open questions about performance with a broader range of whiskies. There’s also the challenge of accounting for how aging in casks influences flavor over time. Peveler noted that external factors, such as the environment where whisky is consumed, also play a significant role in flavor perception and may require further exploration.

Still, it’s interesting to see AI stir up and bring a new layer of precision to an unexpected field like Whisky tasting.