Why is WeChat only developing an input method now? Worries about user profile leaks may be the reason

Last night Zhang Xiaolong said #WeChat will launch its own input method#, and the one doing this is the WeChat artificial intelligence team.

[[377265]]

I know there are quite a few such AI teams within Tencent. They do not serve a specific app, but exist independently as a middleware. When a product within Tencent needs support, they act like a "construction team" to "install" the self-developed functions.

Among the many teams, the most famous ones are Youtu and WeChat AI. The former focuses on image processing and recognition. Examples include several photo-taking and dressing-up apps that have gone viral on WeChat Moments, as well as various graphic applications in Weishi.

WeChat AI does a lot of things, from solving low-resolution QR code recognition to distinguishing fake news and image search. In the early years, it even participated in the development of its own augmented reality framework. This time, Zhang Xiaolong said that the team that would develop the input method for WeChat was this team.

Of course, from the perspective of technical reserves, it should not be a problem to develop an input method. The core of the input method is the algorithm, especially now that major input method companies have optimized AI algorithms, which happens to be what they do, so it gives them an extra bonus.

But this time WeChat suddenly announced that it wanted to develop an input method. Was it just based on concerns about user privacy leaks?

Many years ago, Lei Jun had the honor of participating in the promotion of Baidu's mobile input method. He also wrote the slogan for their poster. During that process, he had some exchanges with them on the direction of mobile input method.

As a program that can be used across multiple platforms and is frequently used, mobile input methods have the ability to aggregate traffic and are a good traffic entrance (the three-stage rocket concept of Wang Xiaochuan, president of Sogou Input Method, is also based on this).

However, compared with the traffic aggregation capability, with the development of AI technology, input methods have been given another powerful application function - the ability to accurately portray user portraits. Not only that, according to different user usage scenarios, more segmented and vertical user feature portrayals can be achieved.

In other words, through the characteristics of your input content in various apps at different times, we can accurately understand your characteristics at different times, thereby helping you "predict the next word" when you input, and even choose a picture that is more suitable for your content at this time.

Do you understand what I mean? Input methods, especially smart input methods, know you much faster than you think. As a company that also relies on advertising and has transformed itself into a weak algorithm for content recommendation, how can WeChat bear to let its competitors "see through me" and my users in this way?