Mackmyra,  a Swedish single malt whisky distillery, says it has announced the world’s first whiskey generated through artificial intelligence (AI).

Working with Fourkind, a Finnish technology consultancy that with expertise in AI, and Microsoft, this is the first time that a complex consumer product recipe has been created with machine learning. The possibilities for different whiskies are endless, says Mackmyra – Master Blenders, who are responsible for the whisky flavour and recipe development, can spend their whole lives tasting and experimenting to create the best flavours possible. However, the distillery’s machine learning models, powered by Microsoft’s Azure cloud platform and AI cognitive services, can analyse existing recipes, sales data and customer preferences to generate more than 70 million recipes that it predicts will be popular and of the highest quality.

Angela D’Orazio, Master Blender at Mackmyra, said “We always strive to challenge the traditions in the very traditional whisky trade and that’s something we really do now when we develop a whisky with the help of AI. We see AI as a part of our digital development, it is really exciting to let AI be a complement to the craft of producing a high-quality whisky. For me as a Master Blender it is a great achievement to be able to say that I’m now also a mentor for the first ever created AI whisky in the world.”

Jarno Kartela, Machine Learning Partner at Fourkind, added“Algorithms don’t have senses so we need another take on how to understand something so complex as whisky. Although lacking human expertise, we can teach machines to understand what elements previous recipes and products are made of and how they are perceived and ranked by customers and experts. With this as a raw data asset, we can leverage a combination of explorative algorithms to generate endless new recipes and products and then use a set of discriminative algorithms to understand which of them might be great, repeating until better recipes are not found. This requires a lot from the computation side, as we need millions of iterations while keeping track of what worked and what did not before reaching a solid guess of a good new whisky.”