When it comes to global expansion, Amazon’s Alexa is taking it slow.
As of now, Amazon devices with the Alexa virtual assistant are only available in three countries, including the U.S. That’s because varying conventions of speech, from place to place, present a unique challenge for natural language processing, as Kintan Brahmbhatt, Amazon Music’s director of product, explained during the Upstream Music Fest + Summit in Seattle on Thursday.
“We need to understand accents,” he said. “We need to understand the local nuances and so on. Simple things, like how customers ask for music, changes drastically between even the U.S. and U.K. U.S. people will ask, ‘hey play the latest single by Green Day.’ In the U.K., they’ll say ‘play that record.'”
Amazon Echo launched in the U.K. and Germany last fall, the first expansion of Alexa-enabled devices into Europe. Brahmbhatt said that Amazon is being careful about expanding into other countries because the company wants the Alexa experience to be “magical” for customers everywhere.
It just goes to show how early-stage natural language processing is. Voice, as an operating system, is in its infancy and the engineers developing the technology are still building the infrastructure necessary to make it work across different dialects and languages.
“If you look at the macro trend for the last 30-40 years, the accuracy of natural language understanding, due to the computing power and all the research that has been done, has been increasing on an exponential scale,” Brahmbhatt said.
Brahmbhatt gave a behind-the-scenes look at Echo and Alexa as part of the Upstream Summit, a series of educational sessions on the intersection of music and technology.
Upstream is a brand-new festival created by Microsoft co-founder Paul Allen. The event kicked off with a keynote from music legend Quincy Jones and will include sessions during the day and musical acts in the evenings through the weekend.
After a deep dive into the technology that powers Echo, Brahmbhatt left his audience with a few Easter eggs from the Alexa engineering team. He encourages anyone with an Echo device to ask Alexa if she can sing in autotune or to explain the difference between “Ice Ice Baby” and “Under Pressure.”