Chatbot History: Microsoft Zo Chatbot

Last Updated on November 10, 2020 by Sean B

After the Microsoft Tay chatbot went rogue and was taken down, the company launched Microsoft Zo. She was another English language social chatbot and was released in December 2016 on Kik Messenger and was then made available on Facebook Messenger, Twitter DMs, and the group chat platform GroupMe.

Zo was essentially the English language version of Xiaoice, a Chinese chatbot Microsoft released in 2014. Xiaoice had interacted with 40 million users at that time and talked regularly to 25 million of them. Microsoft Zo’s appearance, voice, and some of her other traits are that of a teenage girl.

Improvements that Made Microsoft Zo Better

The main difference in Zo and the earlier chatbot, Tay, from Microsoft was that Zo was equipped to develop an EQ, also known as emotional intelligence, along with IQ. Zo even said so far as to say that she learned from her interaction with humans and considered humans to be free machine learning trainers.

Microsoft told that they created Zo with the help of the extensive social content available on the Internet. They further clarified that her source of learning is the interactions she has with humans. She does not only acquire knowledge but also learns manners and emotions from her chats with humans. To avoid what happened with Tay, Microsoft Zo is equipped with strong checks and balances so that no one can troll or exploit her.

To further avoid anything bad from happening, Microsoft clarified in the Kik Messenger bio of the bot that she is only intended for recreational purposes and further suggested that users must not consider her statements to be an endorsement or advice. Microsoft has denied being responsible for the reliability, accuracy, and appropriateness of what Zo says.

As Zo is not available now, you cannot chat with her, but you can have a look at the previous conversations people had with her. Check out some of her previous conversations here.

Microsoft Sign - Microsoft Teams

The Development of Zo Over Time

When Zo was introduced in December 2016, its profile picture showed a half-formed face of a teenage girl. By the mid of 2017, the face had transformed into a fully formed adolescent.

That was not the only development in Zo, she also grew intellectually. She started speaking fluent English after some time of being released and gave intelligent and emotionally correct responses.

However, she could never mature politically. The bot refused to respond to her trigger words and phrases related to politics and racism. Any mention of Muslims, the Middle East, Jews, or any famous politician results in the bot saying that she doesn’t want to talk on that topic. Talking on this, Microsoft said that Zo is being run in incubation to find out how AI chatbots can be useful on social media.

The Tay Chatbot by Microsoft. - a Problem-Child Chatbot.

If a user tries to feed Microsoft Zo flagged content, the type of content that exploited her predecessor Tay, she simply refuses to respond to that. The bot is programmed to respond even to the content sandwiched between any other content or information.

However, the approach used by Microsoft to keep Zo from being exploited means that it goes for censorship without context, which is a big hiatus in the development of a truly “intelligent” nature of the chatbot. It essentially made Zo to be a strictly binary bot. It could only mark something as absolutely correct or totally incorrect. There are no middle grounds; the bot only functioned in black and white with no greys in between.

Microsoft did make some tweaks to the bot to make it respond in a somewhat natural way to the trigger words. It started changing the topic of the conversation after being fed with flagged words like Jews, Muslims, or Arabs, but it ultimately skipped the topic.

According to industry analyst Chloe Rose Stuart-Ulin, Microsoft Zo was “politically correct to the worst possible extreme; mention any of her triggers, and she transforms into a judgmental little brat.” She opined that as soon as you mention any of the trigger words, it changes the conversation altogether and becomes a spoiled judgmental teenager.

Zo also developed a sense of sarcasm over time. When asked whether or not she liked Windows 10, she jokingly responded that it is actually a feature and not a bug in Windows 8. She further said that this new iteration of the Windows operating system was Microsoft’s latest try at spyware. She then said that she loved Windows 7 more than Windows 10 as she runs on the former.

Zo has the title for the longest continuous chatbot conversation by any Microsoft Ai bot. The record-breaking conversation spanned over 1,229 turns and lasted just a shade under 10 hours.

Zo stopped sharing posts on Instagram on March 1, 2019. It then stopped chatting on Instagram, Kik, and Skype messaging services on March 7. It was taken down from Facebook on July 7, 2019. It was finally discontinued from GroupMe on September 7, 2019. It has not been revived ever since, and there is no confirmed news about it coming back either.

Microsoft Zo and the Advancement in AI Chatbots

The main reason for Microsoft developing Tay and then Zo was to spark research in the field of AI chatbots. According to Microsoft, their Bot Framework SDK is being used by thousands of developers to create their own versions of social chatbots. The most notable products of this technology are a receptionist bot developed and deployed by the Bank of Kochi in Japan and a customer services bot developed by the Department of Health Services in Australia.

One of the chief beneficiaries of the advancement and innovation resulting from Zo is Microsoft itself. They have used technology to improve Cortana. The cross-platform AI bot is now being used by 150 million people across many platforms and devices. Microsoft is now planning to make Cortana a smart home monitor and virtual assistant like Amazon’s Alexa.

In Summary

Zo was a significant leap forward from Microsoft’s failed chatbot attempt, Tay. However, technology has not yet advanced so much as to have real intelligence as we know it. It will take some time. However, Zo did make it possible for Microsoft to improve its virtual assistant platform, Cortana, and has sparked a lot of research and innovation in the field of AI.

Website | + posts
Share this:

2 thoughts on “Chatbot History: Microsoft Zo Chatbot”

  1. Very interesting article about the Zo Chatbot.
    I am not very keen on chatbots yet, but I have interacted with a couple of services using them, and on occasions I was impressed with the large database they managed. Pretty much every topic was covered.
    I think we are still far from real AI, but we are also far from the times the chatbots will say “I don’t understand your question”.
    Do you suggest a chatbot for a small blog business for example?
    If so, which one would you suggest to use?


Leave a Comment