AI beats human experts in distinguishing American whiskey from Scotch

Artificial intelligence can tell Scotch whiskey from American whiskey and identify its strongest ingredient aromas more reliably than human experts – using data rather than tasting the drink.

Andreas Graskamp At the Fraunhofer Institute for Process Engineering and Packaging IVV in Germany and colleagues trained an AI molecular odor prediction algorithm called Owsom on descriptions of different whiskeys.

Then, in a study involving 16 samples – nine types of Scotch whiskey and seven types of American bourbon or whiskey – they tasked them with describing drinks from two nations based on keyword descriptions of their flavors, such as floral, fruity, woody or Woody or smoky. Using these alone, the AI ​​could tell which country the person came from with about 94 percent accuracy.

Because the complex aroma of these spirits is determined by the absence or presence of multiple chemical compounds, the researchers also fed the AI ​​a reference dataset of 390 molecules commonly found in whiskey. When they gave the AI ​​data from gas chromatography-mass spectrometry that showed which molecules were present in sample spirits, it increased Ovsum’s ability to distinguish American from Scotch drams by 100 percent.

Compounds like menthol and citronellol were a dead giveaway for American whiskey, while the presence of methyl decanoate and heptanoic acid gave a nod to Scotch.

The researchers tested both Owsom and a neural network on their ability to predict the top five odor keywords based on the chemical content of a whiskey. On a score from 1 for perfect accuracy to 0 for persistent inaccuracy, Ovum achieved 0.72. The neural network achieved 0.78 and human whiskey expert test participants achieved only 0.57.

“(The results) highlight the fact that this is a complex task for humans, but it’s also a complex task for machines – but machines are more consistent than humans,” the team members say. Satnam SinghAlso at the Fraunhofer Institute. “But that’s not to say that humans aren’t needed: We need to train their machines, at least, for now.”

Neither model takes into account the concentration of molecules, only their absence or presence, which is something the researchers hope to improve, and which could also lead to greater accuracy.

Grasskamp says such AI tools could be used for quality control at a distillery, or to help develop new whiskeys, as well as to help detect fraudulent ones. But they can also be used for “anything that smells”, such as in other food and beverage production or the chemical industry.