Human experts spend years roasting and tasting coffee from many places to learn about the beans. They might then combine a coffee bean from Africa with another one from South America to come up with a new blend.
But, a coffee roasting company in Helsinki, Finland is hoping that artificial intelligence, or AI, can help ease the workload.
Kaffa Roastery recently launched its “AI-conic” blend at the Helsinki Coffee Festival. The blend is a mixture of four kinds of beans, including Fazenda Pinhal from Brazil. It is the result of a joint project by Kaffa and Elev, a local AI business.
Elev told The Associated Press that its computer models, similar to ChatGPT and Copilot, created a blend that would “push the boundaries of conventional flavor combinations.”
Svante Hampf is the founder of Kaffa Roastery. He and his partners wanted to do an experiment to see what blend the computer would produce.
“We basically gave descriptions of all our coffee types and their flavors to AI and instructed it to create a new exciting blend,” Hampf said.
The “AI-conic” blend combined beans from Brazil, Colombia, Ethiopia and Guatemala. It is described as “a well-balanced blend of sweetness and ripe fruit.”
Hampf said he thought the combination was “somewhat weird.” He was surprised that the AI program chose to make the blend out of four different kinds of beans. Most humans choose only two or three beans.
However, after the first blind test of the new blend, Kaffa’s coffee experts agreed that the AI-assisted blend was perfect. There was no need to change.
Elev’s Antti Merilehto said “AI-conic” is an example of “how AI can introduce new perspectives to seasoned professionals.”
The International Coffee Organization says that the 5.6 million people in Finland drink the most coffee per person in the world. And the roasters at Kaffa said they hope the AI project will help to bring more things to the strong coffee culture in the country.
Hampf called it a “first step,” adding, “I think AI has plenty to offer us in the long run. We are particularly impressed of the coffee taste descriptions it created.”
I’m Dan Friedell.
