Microsoft also has an AI bot that makes phone calls to humans

Google demonstrated a jaw-dropping new capability in Google Assistant earlier this month, allowing the Assistant to make calls on your behalf. While Google Duplex generated controversy and discussion around artificial intelligence, Microsoft has been testing similar technology with millions of people in China. At an AI event in London today, Microsoft CEO Satya Nadella showed off the company’s Xiaoice (pronounced “SHAO-ICE”) social chat bot.

Microsoft has been testing Xiaoice in China, and Nadella revealed the bot has 500 million “friends” and more than 16 channels for Chinese users to interact with it through WeChat and other popular messaging services. Microsoft has turned Xiaoice, which is Chinese for “little Bing,” into a friendly bot that has convinced some of its users that the bot is a friend or a human being. “Xiaoice has her own TV show, it writes poetry, and it does many interesting things,” reveals Nadella. “It’s a bit of a celebrity.”

While most of Xiaoice’s interactions have been in text conversations, Microsoft has started allowing the chat bot to call people on their phones. It’s not exactly the same as Google Duplex, which uses the Assistant to make calls on your behalf, but instead it holds a phone conversation with you. “One of the things we started doing earlier this year is having full duplex conversations,” explains Nadella. “So now Xiaoice can be conversing with you in WeChat and stop and call you. Then you can just talk to it using voice.” (The term “full duplex” here refers to a conversation where both participants can speak at the same time; it’s not a reference to Google’s product, which was named after the same jargon.)

Nadella demonstrated this functionality in action today, noting that Xiaoice has made a million calls so far. The voice “sounds really good,” according to Verge science reporter Angela Chen, who is fluent in Mandarin. “The pitch is artificially bright, sort of like a US version of a newscaster voice.” Microsoft’s impressive demo showed how its bot can even predict what the person will say next, and respond quickly. In the video demo, Xiaoice interrupts the user mid-sentence to alert them that there are strong winds and they should close the window before bed.

It’s surprising that Microsoft hasn’t demonstrated the same capabilities in its own Cortana digital assistant. Xiaoice is limited to China, and many in the West aren’t even aware it exists despite the chat bot’s fame. Microsoft’s first English-language bot experiment, Tay, ended in a disaster after Twitter users taught it to be racist in less than a day. Microsoft might, understandably, be more cautious about creating another English-speaking bot as a result.

Still, Microsoft clearly has the technology to have bots handling basic phone conversations in China, so it’s arguably only a matter of time before this arrives for users in the US and other English-speaking countries. With Google aiming to start testing Google Duplex in the summer, maybe we’ll all soon be speaking to bots that are reminding us to close our windows and clean our teeth before bed.


There’s a surprise..MS has tech they don’t advertise. They really need to learn to "read the room".

Maybe they are smart enough to test a product first in the Chinese market, before to use a fake demo in a presentation for developer…

Is there proof it was fake?

no there is no proof. they are rumors and some stupid people to spread them

Yeah it’s a rumor until we see some proof. The assistant already understands me extremely well, it just can’t do much with the data. Wouldn’t be surprised if duplex is as good as advertised.

Upon analysis, there were a number of things wrong (or at least inconsistent) with the Google demo call. If the technology is so good, then why didn’t Google do a live demo? Why doesn’t Google answer any follow questions from the media? etc, etc.

Same reason why most don’t do live demos. Just because something unfinished and unreleased works, does not mean all the hiccups have been worked out. Like Ubisoft debuting games that lag and freeze on a stage. Live. Questions won’t be answered until launch, most likely.

Yeah, it’s understood that things can go wrong. Yet, companies often do decide to go with a live demo and quite frankly it results with higher impact and fewer questions. Yes, something could go wrong… but if it’s a technology demo and not even an actual product (like Duplex), who cares? A pre-recorded demo clearly shows a lack of confidence in your own product and it invites questions as to how much of it was edited or orchestrated.

Let me guess: the moon landing was a hoax, too.

I feel that the calls were set up as there is no hairdressing salon I know of (and I know a lot of salons) that does not answer the call with the salon name IE "Welcome to happy cutters, Katie speaking how may I help you". Either way the tec is interesting

In a way true. Just like the pixel earbuds translation. On stage it seemed good, but reality a different story.

To be honest, it’s hard not to have some questions when the Duplex thingy seemingly nails the ‘human sounding voice’ at levels far exceeding anything Google has shown of the Google Assistant.

I’m not going to call it ‘fake’ like the other guy did, but I’d like that cleared up.

Would you be so kind yo find the source of your. Claims?

Would you be so kind yo find the source of your. Claims?

Yeah leaving it up to grainy YouTube videos to leak. Kinda sad. Whoever planned Build 2018 was asleep at the wheel.

The live stage demo of AI at Build was far more impressive and useful than the edited recording played by Google.

I disagree. It didn’t even come close.

Microsoft did a poor job showing AI implementation in their existing products

Microsoft did a poor job showing AI implementation in their existing products

Huh, Microsoft did a live demo onstage.

As far as usefulness you are right, what was shown wasn’t even close. Live dictation of a meeting with multiple participants while automatically generating action items is way more useful to me.
I can’t even remember the last time I called a business for anything.

Different markets and needs for each company. Google is a commercial company where Microsoft is predominantly enterprise these days. I think it’s great they’re both focused on where they can provide the most benefit. As a developer, the majority of meetings I’m in are a waste of time (and I suspect they are for most people). But I think it’s neat that Microsoft is streamlining what little good may come out of them.

People search for business times constantly, and make appointments for haircuts, dental cleanings, yearly physicals, dog groomings, etc.

Please tell me what does it mean to ‘read the room’ in the context of technology like this where there is no ‘room’.

Tay, ended in a disaster

I think it was a resounding success! It managed to assimilate to its environment in record time.

The disaster wasn’t Tay as much as the cesspit we call Twitter.

It had to be a "disaster" or the news blogs wouldn’t have anything to say…

No MS should have put in a morality construct that would not allow it to turn racist.

It didn’t really. It got a copying and reflecting construct that allowed it to mirror racist things. Still fundamentally different from a racist AI. (Still bad, very very bad, but not quite the same problem as you’re saying).

View All Comments
Back to top ↑