The pauses between responses are gone, replaced by natural conversation just like chatting with a friend on the phone.
Personal digital assistants are beginning to fill our homes, most commonly in the form of smart speakers. We’re also using similar assistants online in the form of chatbots. But they all have one thing in common: talking to them isn’t very natural. That’s because these AIs need to listen to what you say, formulate a response, and then talk back. The pauses are what makes it an unnatural exchange.
Microsoft announced this week that it has fixed this problem by developing a chatbot capable of listening and talking at the same time. The end result is much more natural conversations with our AI friends.
Posting on the Microsoft AI Blog, Allison Linn, senior writer at Microsoft, explains how the company has made a technological breakthrough allowing conversations with an AI-powered chatbot to be as natural as speaking to a friend on the phone. The breakthrough happened in China where Microsoft runs a very popular “intelligence-powered social chatbot” called Xiaolce.
Li Zhou, lead engineer on Xiaolce, refers to the breakthrough as allowing for “full duplex voice sense.” So while the bot continues to listen while it talks, the added bonus is it can predict what is going to be said next and take better decisions on how to respond, as well as responding more quickly. Zhou refers to it as “the art of conversation that people use in their daily life,” and it looks as though Microsoft just unlocked that skill for the 200 million+ Xiaolce users.
Xiaolce is already quite advanced, having been designed in order to hold longer conversations with users, tell jokes, partake in chitchat, play games, and now it will do so using a natural conversation style. It doesn’t even need a wake word anymore to activate or continue a conversation.
Although currently limited to Xiaolce, the fact Microsoft seems to have solved the natural conversation problem means it should quickly roll-out to other social AI including Rinna in japan and Zo in the US. Cortana isn’t a social assistant, it’s a productivity assistant, but there’s nothing to say it couldn’t be both. If Microsoft wants to stand out from the crowd, then it should really upgrade Cortana with a natural conversation skill set.