June 18, 2019
At its re:MARS AI conference earlier this month, Amazon previewed Alexa Conversations, a module within the Alexa Skills Kit that melds Alexa voice apps to allow users to perform complicated tasks requiring multiple skills — all with fewer lines of code. That’s because a recurrent neural network will be able to “generate dialogue flow” automatically, thus limiting the number of steps a user needs to order food or reserve a ticket. Amazon vice president David Limp dubbed Conversations “the Holy Grail of voice science.”
VentureBeat reports that Limp defined the Holy Grail as how to “make a conversation string together when you didn’t actually programmatically think about it end-to-end.” “A year or two ago I would have said we didn’t see a way out of that tunnel,” he said, adding that “it will take us years to get more and more conversational.” Still, “this breakthrough is very big for us, tip of the iceberg.”
At re:MARS, a demo showed a woman buying a movie ticket via Atom, make dinner reservations with Open Table and hail a ride with Uber, all early Alexa Conversations partners. With Conversations, the interactions were cut down from 40 to “about a dozen.”
At a gathering of Alexa, Google Assistant, Siri, Cortana and Viv leaders two years ago, Siri co-founder Adam Cheyer noted that the ideal is “one assistant on every device to access every service without any differentiation between what’s core and what’s third-party.”
That’s the path Amazon is intending to follow, starting with last fall’s introduction of Follow-Up Mode, which allows the user to “engage in multiple interactions but only have to say the ‘Alexa’ wake word once.”
Google is also attempting to make its assistant “capable of fluid conversation,” debuting Continued Conversation so that users “don’t have the say the wake word to continue to talk about something.” Microsoft purchased startup Semantic Machines in 2018.
“When you’re starting to solve more complex problems, there is more give and take of information, there are more decisions at each point in time, and hence there are multiple actions that can come in context of the same conversation with different individuals,” said Alexa AI senior product manager Sanju Pancholi.
Pacholi noted, “If you make [customers] repeat information again and again and again, you are forcing them to believe that they are talking to a dumb entity, and if that’s the rapport you’re building with them from the get-go, the chances are they’re never going to delegate higher order tasks to you, because they will never think you’re capable of solving higher-order problems for them.”
The next step for Alexa Conversations “may include collections of skills to help people watch content at home, get food delivered, or buy a gift,” although Alexa chief scientist Rohit Prasad “declined to share details about use cases that may be taken up next, but believes this could include ways to help plan a weekend.”
Conversations may soon be incorporated in Blueprints, voice app templates for private custom Echo skills, he added.
For more information on Alexa Conversations, visit the Amazon developers site.