German researchers develop AI that’s better than humans at identifying liquors

The world of whiskey authentication and analysis is now also being transformed by artificial intelligence, with new research showing AI systems can outperform human experts in distinguishing between American whiskey and Scotch.

Breakthrough findings: Researchers at Germany’s Fraunhofer Institute have developed an AI algorithm called OWSum that achieves unprecedented accuracy in whiskey classification.

• The AI system demonstrated 94% accuracy in identifying whiskey origin using only flavor descriptions
• When analyzing chemical data through gas chromatography-mass spectrometry, the system achieved perfect accuracy
• Human whiskey experts were significantly outperformed by the AI, scoring 0.57 compared to the AI’s 0.72-0.78 on odor prediction tasks

Technical methodology: The research team identified specific chemical compounds that serve as reliable markers for distinguishing between American and Scottish whiskeys.

• Menthol and citronellol were found to be key indicators of American whiskey
• Methyl decanoate and heptanoic acid emerged as distinctive markers for Scotch
• The current AI models analyze the presence of molecules but do not yet account for concentration levels

Industry implications: The technology shows promising applications across the spirits industry and beyond.

• Distilleries could implement the system for enhanced quality control measures
• The AI could assist in new whiskey development and innovation
• The technology offers potential for detecting counterfeit products and fraud in the spirits market
• Similar applications could extend to other industries where molecular analysis and scent identification are crucial

Research context: The study, published in Nature Communications Chemistry, represents a significant advancement in applying machine learning to beverage analysis.

• The research was led by Andreas Grasskamp and his team at the Fraunhofer Institute for Process Engineering and Packaging IVV
• The approach combines traditional chemical analysis with cutting-edge AI technology
• The system’s success demonstrates the potential for AI to enhance traditional expert-based evaluation methods

Future developments: While the current system shows impressive capabilities, several areas remain for potential advancement and broader application.

• Incorporating molecular concentration analysis could further improve accuracy
• The technology could be adapted for other types of spirits and beverages
• Integration with existing quality control systems could streamline production processes