Chatbots are sort of like Siri or Cortana. They don't have as strong a focus on productivity that personal assistants do. They're designed specifically for longer conversations, to humor users and, at their core, to act as proxy for humans.
They're not usually tied to individual devices like the aforementioned personal assistants, but the end-purpose of "talking" to a user to figure out what they need is common between the two, leading chatbots to be found often as a first line of defense between a customer and an actual human responding to a call if the chatbot's efforts fails to satisfy.
Unlike the chatbots of yesteryears, which generally followed a pre-written script thereby limiting the number of topics they could tackle on their own, Microsoft's are among the newer batch, using artificial intelligence to make small talk instead, drawing their data over time from the people who end up talking to it.
What no one realized at the time was that a few of the people who frequent Facebook and Twitter - platforms where Microsoft's chatbots are available to be spoken with - aren't particularly known to exhibit moral restraint. Given the moral compasses of the chatbots in question are literally shaped in part by internet trolls, what we got in the end were hilariously dark outcomes.
That aside, while these bots are certainly fun to talk to in some ways and are capable of replying to messages based on context, they're not exactly what one would call conversational. They're limited to replies alone, and aren't capable of initiating a new conversation on their own in ways one would expect a human to.
This is the caveat Microsoft's Li Zhou aims to tackle. Zhou is the engineer lead for Xiaolce - often seen as Cortana's China-based baby sister - and he believes that Microsoft's latest breakthroughs pertinent to more human-sounding chatbots could make conversations with them more akin to "talking on the phone to a friend".
According to Zhou, these 'breakthroughs' - already incorporated into Xiaolce - will allow her to converse in "full-duplex", which is essentially tech-speak for 'two-way conversation', unlike the previous walkie-talkie like "half-duplex" interactions. Full-duplex will also serve to reduce the time it takes for a bot to respond, further naturalizing conversation between man and machine.
Di Li, Xiaolce's general manager, says that the end-goal for Microsoft's chatbots is for them to "understand people's emotional as well as intellectual needs".
Microsoft plans to extend this new feature to its other chatbots, including the millenial-oriented Zo, the Indian Ruuh and the Japanese/Indonesian Rinna.
Source: Microsoft Blog
8 Comments - Add comment