Alexa’s New Skill Resumption and Name-Free Interaction Simulate Human Memory and Inference Ability
Amazon is making Alexa more proactive in finding and running voice skills thanks to the combination of its new name-free interaction (NFI) toolkit and Skill Resumption feature. The two new facets for the voice assistant are both available in limited preview, just a few of the torrent of updates announced at the Alexa Live virtual event on Wednesday such as Alexa for Apps.
No Name Needed
Currently, Alexa users need to remember the name of every skill they wish to use right now. The NFI toolkit is designed to shift that burden to Alexa. Instead of citing a specific skill name, users would theoretically only need to use keywords when asking Alexa to carry out a task. Like in human conversations, Alexa would then infer what skill it should access, such as Uber when someone asks the voice assistant to get them a ride, or a recently played game if the request is just to play a game. For developers, there is an additional option where they can pick as many as five potential phrases that customers might say when indirectly referencing their skill, and Alexa will then know it is that voice app being obliquely referenced by a user. If it isn’t completely certain, the voice assistant will ask additional questions of the user to make sure the right skill is opened. As the AI learns what skills a customer prefers, it will become more accurate in choosing which to open.
The NFI toolkit extends the nameless invocation system Amazon set up in 2018 in some ways. There is more flexibility and personalization now, but it still works with the CanFulfillIntentRequest (CFIR) from then to try and limit pulling up an irrelevant skill that happens to have the same keywords. According to Amazon, it’s already paying dividends for developers with some reporting as much as 15% more traffic in the beta.
Pause and Resume
Much as the NFI toolkit can enable Alexa to appear to mimic human inference, the new Skill Resumption feature creates an illusion of the memory and multi-track thinking people do every day in conversation. Instead of only performing one task at a time and needing to finish it before moving on to the next one, Alexa can instead pause a task, running it in the background until it becomes relevant again while handling other requests. In the transportation scenario, Alexa would handle a request for a ride through Uber, then alert the user about any updates on when it will arrive, or answer the same kinds of questions if specifically asked. And in another similarity to the NFI, the user doesn’t need to mention the skill by name, asking where their ride is is enough. While awaiting the car, the user can play music or use other voice apps without missing any updates from Uber.
Skill Resumption is a feature common to smartphones, where music can be interrupted by a phone call and resume after the call ends, or a navigation app can run in the background while a podcast plays, interrupting the audio to give new directions as needed. But, when added to the NFI options, Alexa could simulate a human conversational partner far better than it does right now. Being able to rely on the AI’s version of memory and ability to infer from context and history what someone is talking about might eliminate a lot of the friction of using a voice assistant, overcoming the reluctance of some who don’t want to have to learn the special etiquette to get Alexa to do what they want.