You may be used to asking Alexa to play music, tell you the weather, or any number of simple requests. But what if you commanded Amazon’s voice assistant to buy movie tickets, and it asked if you also want a restaurant reservation before the show?
That’s the near future Rohit Prasad, vice president and head scientist for Alexa, unveiled Wednesday at Amazon’s re:MARS artificial intelligence conference in Las Vegas. It plans to release the feature in the coming months. Amazon is working with a number of companies on the experience, including Atom Tickets, ride-hailing company Uber, and restaurant reservation service OpenTable.
The market for voice-controlled assistants is booming, and Amazon — along with competitors such as Google and Apple — wants people to interact with its in increasingly natural-sounding ways. This includes letting them speak the way they normally would, shifting from one topic to another with ease, and using the assistant to make decisions without a lot of back-and-forth.
While Alexa has more than 90,000 voice applications, which Amazon refers to as “skills,” they’re typically used one at a time. Yet a videotaped demonstration showed how Alexa could soon be able to arrange a whole evening out for you — something Google’s Assistant and Apple’s Siri can’t yet do.
First, a woman’s voice asked Alexa for show times for the film “Dark Phoenix” on Saturday night. After some back and forth, the woman settled on a time, and asked Alexa to purchase two tickets to a 7:05 p.m. show at Regal Cinebarre Palace Station, a movie theater in Las Vegas.
Then, Alexa did something new, asking the woman, “Will you be eating out near Regal Cinebarre Palace Station?”
The woman requested a Chinese restaurant, and Alexa offered several suggestions. The woman asked Alexa to make a reservation at one of them. After that, Alexa asked if she needed an Uber ride to the restaurant.
Once Alexa booked an Uber, it asked, “Anything else?”
“Show me the trailer,” the woman said, referring to a preview for the upcoming movie.
Alexa complied, playing the trailer for “Dark Phoenix” on the Echo Show’s display.
Today, you’d have to talk to Alexa 40 times to complete this kind of complex interaction, Prasad said. But the feature he demonstrated required only 13 user turns and far less time, he said.
Alexa is already somewhat proactive with features like Alexa Guard, which lets Amazon Echo devices listen for suspicious sounds such as breaking glass. And you can say things like “Alexa, turn on the lights and play Rihanna” to get the assistant to do two things at once. But it’s a lot trickier to create the kinds of dinner-and-a-movie (plus an Uber ride) scenarios that Amazon envisions.
Making this kind of complicated interaction work requires a lot of AI effort — Alexa needs to not only understand and respond to human queries, but also come up with smart suggestions that could add to the conversation. Speaking on stage at the conference, Prasad said this works by using a cross-skill action predictor — essentially, software that can reply to commands like “get me movie tickets” by getting the movie tickets and be proactive in pulling up other skills like reserving a table at a restaurant.
Talking to reporters at re:MARS after the morning’s keynote talks, Dave Limp, senior vice president of Amazon devices and services, said the company is working on a couple other experiences beyond the night-out scenario that will also roll out in the coming months.
“It will take us years to get more and more conversational, but this breakthrough is very big for us,” he said. “Tip of the iceberg.”