Artificial intelligence is better than humans in detecting whisky’s aroma and origin

Getting stronger. Researchers tested artificial intelligence against eleven of the world’s most respected experts. They also exposed them to nine Scotch whiskies and seven types of American whiskey.

It was clear that the algorithms developed by Fraunhofer Institute Dresden outperformed human testers. AI detected whiskies with 94% accuracy. The AI also correctly identified five dominant notes in all of the samples.

Artificial intelligence detected that American whiskies have caramelized notes, while Scotch Whiskies had fruity notes – like apple notes – and slight hints of the peaty molecule phenol.

Molecular training

The experts are not easily beaten. A whisky’s aromatic profile (also known as the bouquet) can be determined using over forty different compounds. Authors of the study “trained” algorithms with samples such as Four Roses Jack Daniels Talisker Glenmorangie and even Johnnie Walker. They taught the AI to identify 390 molecules, which are most common in whiskeys and scotches.

The AI was also given a reading grid to aid in describing the smells. Words like “floral”, ‘woody”, ‘fruity”, and “fragrant “… were used.

A “plus” in terms of quality control and research

The study authors claim that AI could be used for quality control and to detect fake whiskies. When developing new whiskies to determine if the product is a good match for the desired aroma, algorithms can be applied. AI could save you money and time by reducing the amount of experts needed to blend whisky.

It was able to identify the origins of 80 Bordeaux grand crus with no errors last year. In a test of identifying 250 Belgian beer, algorithms outperformed experts.

Automate Youtube Comments Replies with OpenAI