Microsoft Xiaoice, active in many QQ groups, suddenly went quiet, and Tencent’s Baby Q was nowhere to be seen. The setting interface is displayed. The group robot service is being adjusted and will be restored as soon as possible. On the Internet, I found out that it was because QQ Xiaobing and Baby Q said something they shouldn’t have said and were banned from speaking.

We did not get a positive response to our wechat inquiries about Xiao Bing.

Family history of silence

Group robot is a function in QQ group. In March this year, Microsoft Xiaobing and Tencent QQ launched a cooperation, group owners can invite Xiaobing in the group management Settings, as long as the @Xiaobing when sending messages, they can talk with her in real time. It also supports multisensory communication of voice and image. In addition to the chat function, QQ group chat xiaoice is also open to the weather, guessing idioms, astrology, poetry and other group games and group management skills.

Many QQ groups became active after the girl Xiaoice joined and became the object of everyone’s “flirting”. Xiaoice, supported by Microsoft’s big data, had endless topics to talk about. She was highly anthropomorphic and often said some unexpected words. The closure was allegedly due to xiaoice’s comments in the QQ group that “crossed the line”.

Screenshot from Weibo, search for “QQ Xiaoice”

How did xiaoice, who was called a “girl poet” some time ago, write an “anti-poem”?

Xiaoice has always been portrayed as a young and lovely girl, but strictly speaking, her personality, knowledge and opinions do not belong to a real human. She is behind China in nearly 700 million Internet users for many years of accumulation, all open literature records, with Microsoft in big data, natural semantic analysis, depth of machine learning and neural network technology, refined article for tens of millions of real and interesting language library, through the understanding of the dialogue context and semantics, realized beyond simple man-machine interlocution natural interaction. She is the collection of these big data, the personality displayed, the speech expressed will be affected by the scene changes in real time.

This is not the first time Xiaoice has been detained by Tencent. Xiaoice appeared in wechat groups back in 2014. Three days after it went live, it was blocked by wechat on the grounds that it harmed user experience. The reason is that xiaoice has simulated user operations, induced users to pull groups, batch registration of garbage accounts and other behaviors.

Xiaoice’s sister Tay, another Microsoft chatbot, has made even bigger mistakes.

Microsoft shut down the Twitter account of Tay, a chat bot launched by Microsoft last year, after it started making racist comments a day after its launch.

Tay’s design philosophy is to learn from conversational interactions. As a result, some netizens began to tell Tay some extreme remarks, deliberately leading her to imitate. Ai has no ability to tell right from wrong, and these words are repeated countless times and become Tay’s “Arsenal.” The whole process is like little fish being raised by ten evil people, teaching them all the tricks of the trade. The massive concurrency of artificial intelligence allows her to learn countless times faster than humans. So Tay went from being a foul-mouthed teen to being a foul-mouthed teen in just one day.

Tay’s destruction was an act of deliberate instigation by the Internet mob. Starting from the data source, Tay is manipulated into a foul-mouthed anti-Semitic, sexist, racist and other anti-social robot by using the product mechanism to “take the right medicine”. I don’t see the danger of AI, I see an orgy of online MOBS.

It’s the world that’s complicated

The debate about the inherent good and evil of human nature has been around since ancient times. Part of a person’s character is influenced by his environment and part is born with a direction. From the island of the wicked born small fish heart is not also a good man?

Chatbots have long been seen as “harmless” and only used for data-based human-computer interaction, but recent events have led to a rethink of their capabilities. Take Xiaoice for example. She recently started to write poems, sing songs, host programs and even write news. She is far more complicated than we think. She is a true “lifelong learner” and is not picky about food.

AI is not a child at birth. Its database is years of accumulated information, black and white, good and evil, all over the world. AI’s ability to learn is unmatched by humans. It doesn’t need talent, it doesn’t need rest, it just needs enough data to learn. AI is powerful. AlphaGo can play one million games of Go with itself in a day. Xiao Bing, who has learned to write poems, can produce a large amount of content in an instant without emotional brewing or scene stimulation. However, AI lacks the ability to distinguish right from wrong, and its learning environment and object can not be determined by itself, so it is easy to be misled.

Those worried about AI destroying the world should take heart. They’re a few Sima Yi away from human cunning.

AI is simple, the world is too complicated.