A new study reveals that artificial intelligence algorithms are better at distinguishing whiskey aromas than humans.
The results of the research, shared in Communications Chemistry last Thursday (19), show how chemists were able to create two machine learning algorithms, OWSum and CNN.
The first is a statistical tool for perceiving molecular odors, while the other is a convolutional neural network that helps to discover relationships in very complex datasets, explained Andreas Grasskamp, a researcher at the Fraunhofer Institute for Process Engineering and Packaging IVV in Freising, Germany, and the lead author of the study, to AFP.
After training with a list of molecules detected by gas chromatography and mass spectrometry from 16 whiskey samples, OWSum was able to determine whether a whiskey was American or Scottish with over 90% accuracy.
In the second stage, both algorithms were tested to see if they could predict the olfactory qualities of the whiskeys based on the detected molecules or their structural characteristics.
Both OWSum and CNN were able to identify the five dominant notes of a whiskey more accurately and consistently, on average, than any human expert on the panel.
“We found that our algorithms aligned better with the panel results than each individual member, providing a better estimate of the overall odor perception,” explained Grasskamp.
Photo and video: Unsplash. This content was created with the help of AI and reviewed by the editorial team.
