Artificial intelligence surpasses humans in predicting the aroma of a whiskey

What if the artificial intelligence Will it surpass the human being in the art of choosing a single malt? Machine learning algorithms have managed to predict the dominant aromas of different whiskeys better than an expert, according to a study published Thursday.

In our environment, most odors are made up of a complex mixture of molecules that interact in our olfactory system to create a specific impression.

This is the case of whiskey, whose aromatic profile can be determined from more than 40 compounds and may contain an even greater number of non-odorous volatile compounds.

This makes it particularly difficult to evaluate or predict the aromatic characteristics of a whiskey based solely on its molecular composition.

However, This is precisely what chemists managed to do thanks to two machine learning algorithms according to the results of a study published Thursday in Communications Chemistry.

The first algorithm, OWSum, is a statistical tool for predicting molecular odors developed by the study authors.

The second, CNN, is a convolutional neural network that helps discover relationships in very complex data sets, such as those between “the most influential molecules and aroma attributes” in a blended whiskey.Andreas Grasskamp, ​​researcher at the Fraunhofer Institute for Process Engineering and Packaging IVV in Freising (Germany) and lead author of the study, explained to AFP.

The researchers “trained” the algorithms by providing them with a list of molecules detected by gas chromatography-mass spectrometry in 16 whiskey samples: Talisker Isle of Skye Malt (10 years old), Glenmorangie Original, Four Roses Single Barrel, Johnnie Walker Red Label or Jack Daniel’s, among others.

They were also given the aroma descriptors determined for each sample by a panel of 11 experts.

The algorithms were then used to identify each whiskey’s country of origin and its five dominant notes.

Counterfeit Detection

OWSum managed to determine whether a whiskey was American or Scottish with greater than 90% accuracy.

Detection of compounds such as menthol and citronellol were strongly associated with a classification as an American whiskey, while detection of methyl decanoate and heptanoic acid were primarily linked to Scotch whiskeys.

The algorithm also identified caramelized notes as the most characteristic of American whiskeys, while “apple,” “solvent,” and “phenolic” notes (often described as a smoky or medicinal aroma) were the most characteristic of the whiskeys. Scots.

In a second stage, the researchers asked OWSum and CNN to predict the olfactory qualities of the whiskeys based on the molecules detected or their structural characteristics.

Both algorithms managed to identify the five dominant notes of a whiskey with greater precision and consistency, on average, than any human expert on the panel.

“We found that our algorithms align better with the panel results than each individual member, thus providing a better estimate of overall odor perception,” Grasskamp noted.

These machine learning methodologies could be used to detect counterfeits or to evaluate whether blended whiskey “will have the expected aroma, thus helping to reduce costs by limiting the need for evaluation panels,” he estimates.

Could similar results be obtained with wine? “In theory, yes. All these tools need is a list of compounds detected in the sample and their corresponding descriptors”according to Grasskamp.

“The challenge lies in the finer details, such as determining whether wine aromas are distinctive enough for an AI algorithm,” he added.