'Natural Language,' App Linking Highlight Amazon's Next Stage of Voice Control

Amazon’s Alexa Live 2020 voice developer event kicked off virtually Wednesday with announcements of new experiences involving conversational artificial intelligence, graphics and multimedia, on-the-go interaction and skills discovery. Alexa engagements have quadrupled over the past two years, said Aaron Rubenson, vice president-Alexa Voice Service and skills. He said customers interact with Alexa “billions of times each week.”

Rubenson said 90% percent of Alexa-enabled devices last year were built outside of Amazon in TVs, speakers, thermostats, lights, garage doors, locks, headphones and smartwatches, among others. There are more than 100,000 Alexa-compatible smart home products from 9,500 brands.

Last year, Amazon launched Amazon Connect Kit, which allows device makers to make any product with power into an Alexa-compatible smart device using an ACK module. Using ACK, Amazon took the first Alexa-connected product “from idea to mass manufacturing in a matter of weeks.” A new module, announced Wednesday, is 50% lower in cost, Rubenson said, and could add Alexa capability for as little as $4 per device.

Rubenson imagined a future of gaming with Alexa, including visuals on a Fire TV. “Imagine Alexa as a game host that can guide the game and keep track of points for each player,” he said. “Imagine an experience where you feel like you’re in an actual game show.” To get there, developers will have to build experiences allowing customers to speak naturally and handle conversational design patterns, he said.

At the event, Amazon is previewing Alexa for Apps, which lets developers take customers into iOS and Android mobile apps so they can make Alexa skill requests using the Alexa app on phones or accessories such as Echo Buds and Bose with Alexa built-in headphones, said Amazon’s Nedim Fresko, vice president-devices and developers.

Consumers can use Alexa to search, view more information and access features inside the app, he said. Alexa for Apps can be implemented with any app that can be opened with deep links and is being added to experiences for TikTok, Yellow Pages, Uber, Sonic, Zynga and Volley, Fresko said. Alexa for Apps allows developers to give customers additional visual information. After customers book a ride by voice with the Uber skill, Alexa can ask them if they want to see the driver’s location on a map in the app, he said.

He also announced the name-free interaction (NFI) toolkit, designed to make it easier for customers to find and open skills without having to remember the skills' names. Alexa uses keywords, descriptions and categories to surface relevant skills when customers say things like “play an adventure game,” he said.

Amazon is previewing skill resumption in the Alexa Skills Kit, enabling customers to easily return to a skill after a short break or after using Alexa to complete a separate task, said Fresko. A user could, say, “Alexa, where's my ride?” and Alexa will remember that the customer was using Uber, eliminating the need to invoke the skill by name or start over, he said. After confirming the location and destination, the Uber skill moves to the background while matching the customer with a driver and processing the ride request, he said.

Alexa Conversation, meanwhile, in beta, helps create more “natural” Alexa skills with fewer lines of code, using an AI-driven approach to dialogue management, said Fresko. Conversations are more nuanced than simply understanding words and sentences, and developers need to accommodate a broad range of phrases and unexpected dialogue turns -- with enough memory to sustain long back-and-forth sessions, he said.

Fresko gave the example of ordering a pizza by voice and being able to accommodate order details for size, number and type of toppings, how many people it feeds, an order correction such as changing size and applying a coupon. Supporting conversational patterns make Alexa skills feel more natural to customers, he said. A pizza ordering skill with seven topping combinations could require more than 5,000 dialogue paths.

In early product design work with iRobot, Amazon enhanced the customer experience of the robot vacuum’s Alexa skill beyond just scheduling a clean, said Fresko. Using Alexa Conversation, customers can follow several different conversation paths vs. rigid commands, including being able to make changes without starting over and to speak more naturally, he said.

New personalization features in app-to-app account linking allow customers to enable a skill by linking their Alexa account to a service in a two-touch process, Fresko said. A pizza delivery service could send text updates of a food order, he said.

Amazon is previewing an experience called Limit Access, which developers can use to verify which voice profile is accessing their skill using two-factor authentication requiring a passcode, he said. That allows customers to link accounts without having to enter third-party account credentials.

Dallas Mavericks owner Mark Cuban had a cameo during the virtual event, describing different ways he uses Alexa at the home and office. His family has a safety protocol that uses the voice assistant, he said. Saying “Alexa, fire” triggers a script telling family what to do in case of a fire emergency, he said.