In a groundbreaking study, artificial intelligence has outperformed human experts in identifying the differences between American whiskey and Scotch whisky.
Researchers at the Fraunhofer Institute for Process Engineering and Packaging IVV in Germany developed an AI model that achieved remarkable accuracy by analysing flavour profiles and chemical data, marking a significant advancement in the field of sensory analysis.
Key Takeaways
AI model OWSum achieved 100% accuracy in distinguishing between American whiskey and Scotch when using chemical data.
The study involved 16 samples: 9 Scotch whiskies and 7 American whiskeys.
Human experts scored significantly lower than the AI in identifying flavour characteristics.
The Study's Methodology
The research team, led by Andreas Grasskamp, trained a molecular odour prediction algorithm named OWSum. This AI model was fed descriptions of various whiskies, focusing on their unique flavour profiles. The study involved:
Sample Selection: 16 whisky samples were chosen, including 9 types of Scotch and 7 types of American whiskey.
Data Input: OWSum was trained on keyword descriptions of flavours such as flowery, fruity, woody, and smoky.
Chemical Analysis: The AI was also provided with a dataset of 390 molecules commonly found in whiskies, enhancing its ability to differentiate between the two types of spirits.
Results of the Experiment
The results were striking. When relying solely on flavour descriptions, OWSum achieved an impressive accuracy rate of nearly 94%. However, when the AI was given chemical data from gas chromatography–mass spectrometry, its accuracy soared to 100%.
Key Chemical Indicators:
American Whiskey: Identified by compounds like menthol and citronellol.
Scotch Whisky: Recognised through the presence of methyl decanoate and heptanoic acid.
In a comparative analysis, OWSum scored 0.72 in predicting the top five odour keywords based on chemical content, while a neural network achieved 0.78. In contrast, human experts only managed a score of 0.57, highlighting the challenges they face in sensory evaluation.
Implications for the Industry
The implications of this research are significant for the whisky industry. The ability of AI to consistently and accurately identify different types of whisky could revolutionise quality control processes in distilleries. Potential applications include:
Quality Control: Ensuring that products meet specific flavour profiles.
Product Development: Assisting in the creation of new whisky blends by predicting flavour outcomes.
Fraud Detection: Identifying counterfeit products based on chemical signatures.
The Future of AI in Sensory Analysis
While the results are promising, the researchers acknowledge that neither the AI model nor the neural network currently accounts for the concentration of molecules, which could further enhance accuracy. Grasskamp suggests that AI tools could extend beyond whisky to other areas of food and drink production, as well as the chemical industry, where scent plays a crucial role.
As AI continues to evolve, its role in sensory analysis may redefine how we understand and appreciate the complexities of flavour, challenging the traditional expertise of human tasters. This study not only showcases the capabilities of AI but also raises questions about the future of human expertise in sensory evaluation.
Sources
AI beats human experts at distinguishing American whiskey from Scotch | New Scientist, New Scientist.
A.I. Beat Experts In A Blind Taste Test Identifying Whiskey Flavors, BroBible.
