Scientists teach AI to distinguish between different whisky aromas

Researchers have successfully utilized artificial intelligence (AI) to predict the aromatic notes of whisky and determine whether a sample originates from the United States or Scotland. This innovative approach represents a step toward automated systems capable of assessing the complex aromas of whisky based on its molecular composition. 

Dr. Andreas Grasskamp, who led the research at the Fraunhofer Institute for Process Engineering and Packaging in Freising, Germany, emphasized that AI offers unmatched consistency compared to human panels. “We are not replacing the human nose with this,” Grasskamp explained. “We are supporting it through efficiency and consistency,” The Guardian cited.

Traditionally, expert panels evaluate whisky for woody, smoky, buttery, or caramel aromas to ensure consistency across batches. Whisky’s aroma arises from a complex interplay of chemicals that interact within the nose, often masking one another to produce a unique sensory experience. These complexities make it challenging to predict a whisky’s smell solely from its chemical signature. For this study, researchers analyzed the molecular profiles of 16 whiskies, including popular US brands like Jack Daniel’s and Maker’s Mark, and Scottish labels like Laphroaig and Talisker. They also incorporated aroma data provided by an 11-member expert panel to train AI algorithms.

The AI achieved an impressive 90% accuracy in distinguishing between US and Scottish whiskies. It also consistently identified the five strongest aromatic notes of each whisky, outperforming individual panelists in accuracy and consistency. However, researchers noted that the algorithm's performance might decline when analyzing whiskies it wasn’t trained on. The findings were published in Communications Chemistry.

Specific chemical compounds played a crucial role in identifying the whiskies’ origins. For instance, menthol and citronellol were markers for US whiskies, known for their caramel-like notes. In contrast, methyl decanoate and heptanoic acid were linked to the smoky or medicinal aromas characteristic of Scotch whiskies.

Beyond whisky, this AI-driven approach has potential applications in diverse fields. It could be used to detect counterfeit products by identifying discrepancies in their smell or to improve the recycling process by blending old plastics into new products without noticeable odors. 

Dr. William Peveler, a senior lecturer in chemistry at the University of Glasgow, praised the method for its potential to provide greater stability than human panels. Peveler also highlighted the subjective nature of whisky tasting, influenced by environmental factors and personal perceptions. These variables present challenges for AI to fully replicate the human experience of flavor. Despite this, the research marks a significant step toward integrating AI into industries where sensory evaluation plays a critical role.

By Nazrin Sadigova