An artificial intelligence algorithm called OWSum developed by Germany's Fraunhofer Institute can distinguish between Scotch whiskey and American whiskey with near-perfect accuracy, even surpassing that of human experts. By analyzing the whiskey's descriptive keywords (such as floral, fruity, etc.) and chemical components, the algorithm successfully identified the key compounds that distinguish the two types of whiskey. This research not only demonstrates the potential of AI in the field of sensory analysis, but also provides new technical means for breweries’ quality control, new product development, and combating counterfeit products.
Recently, a research team from the Fraunhofer Institute for Process Engineering and Packaging in Germany developed an artificial intelligence molecular odor prediction algorithm called OWSum, which successfully distinguished American whiskey from Scotch whiskey, and its accuracy exceeded Human experts. The team used whiskey flavor descriptions and chemical data to train the AI tool and explore its potential for whiskey identification.
Picture source note: The picture is generated by AI, and the picture authorization service provider Midjourney
For the study, the researchers selected 16 samples, including nine Scotch whiskeys and seven American bourbons. By analyzing the keyword descriptions of these whiskeys, such as floral, fruity, woody, and smoky, OWSum was able to distinguish between the two types of whiskey with an accuracy of nearly 94%. As the research deepened, the research team further provided AI with a reference data set containing 390 common whiskey molecules. Combined with the results of gas chromatography-mass spectrometry analysis, OWSum's discrimination accuracy increased to 100%.
Using the data, the researchers found that certain compounds such as menthol and citronellol were distinctly characteristic of American whiskey, while methyl decanoate and enanthate were more commonly found in Scotch whiskey. In addition, the research team also tested the ability of OWSum and neural networks to predict the main odor keywords of whiskey based on chemical components. In this test, OWSum scored 0.72, while the neural network scored 0.78, while the human expert scored only 0.57. This shows that while AI excels at such tasks, discerning the complexity of whiskey remains a challenge for humans.
Research member Satnam Singh said that although machines perform more consistently, humans still play an important role in training machines. In the future, the research team hopes to improve the model so that it can take into account the concentration of the compound, thereby further improving accuracy. Grasskamp said such AI tools could be used not only for quality control in distilleries, but also to help develop new whiskeys and identify counterfeit products. In addition, this technology has the potential to be used in other areas such as food and beverage production and the chemical industry.
Paper: https://www.nature.com/articles/s42004-024-01373-2
Highlight:
Artificial intelligence OWSum distinguishes American whiskey from Scotch whiskey with nearly 100% accuracy, surpassing the performance of human experts.
The AI analyzes whiskey's odor keywords and chemical composition to identify specific compounds that differentiate the two spirits.
Humans are still important in machine training, and in the future we hope to improve the accuracy of the model to handle more complex tasks.
The successful application of the OWSum algorithm indicates that artificial intelligence technology has broad application prospects in the food and beverage industry, and may promote technological innovation in more related fields in the future. This research result not only improves the accuracy of whiskey identification, but also provides new ideas and methods for other sensory analysis fields.