On August 15, 2011, Lu Zhou's notes "Does it as a last resort":

It was a very difficult decision, but as before, I had to press this button.

At this time last year, the scale of WeChat and Mengguyun had not reached such a large scale. Strictly speaking, only Menggu advertising's business entered the commercial stage. Zhou Wen was responsible for the data processing of Menggu advertisements at that time, and this became the prototype of Mengguyun's later D+.

Of course, in the future development, even Menggu Advertising has moved related business data processing to Mengguyun's d+.

Everything worked smoothly. Menggu Advertising recruited a group of data staff for this purpose, and internally called data engineers. They process data through the visual graphics or data interface provided by d+. This should be called empirical science, which relies on the engineer's experience and knowledge.

At the beginning of WeChat, Cheng Xuyuan did not over-think the problem. As it turned out, this was a wise choice. Under the attack of the penguin, I thought that promoting the iteration of the product was higher than optimizing the data. Before the Spring Festival this year, when data processing was put on the agenda of WeChat, everything became almost out of control.

It is obvious that even if WeChat recruits thousands of data engineers, it will never be able to complete the data processing. Because as the data volume increases exponentially, the amount of data to be processed will always remain above 90%. The manpower is limited after all, but the data is infinite.

Therefore, artificial intelligence is inevitable. It can replace more manpower.

To remove the whole thing from some concepts, I think there are only three: data, algorithms, and computing power. Even if you don’t talk about artificial intelligence, I think you still need to face these three problems.

Data is constantly expanding, whether it is WeChat or other companies. Its size is so huge that it is a bit unimaginable to me. And it is still producing a large amount every day, constantly consuming the storage hardware purchased by Mengguyun.

Most algorithms have appeared in the last century, and what we do is to reconstruct and engineer it. This is a problem that Seven Stars Wolong, WeXun and Mengguyun are discussing together.

Computing power, computing power. On the road of evolution, it has also reached the point where it has to be considered. The CPU of general-purpose computing is destined to be a relatively poor choice in this field. Just as I saw someone discussing FPGA mining machines in the Bitcoin community, for artificial intelligence and data processing within the company, I think higher-level hardware is also needed, such as FPGA and Asic.

In terms of hardware evolution, artificial intelligence and blockchain are somewhat subtle similar. Of course, the correlation is not as great as I imagined. Or is it still difficult for me to see it from afar? Maybe.

After sorting out this, the following discussion will not be too difficult. I divided it into several levels: application platform, algorithm implementation, framework interface and hardware.

Hardware machines are the general component of computing power and are the physical foundation; the framework interface realizes the channel between hardware and software, exposing the interface communicating with hardware to the upper layer for use; the algorithm is implemented based on existing theories, and we may try to do basic recognition of graphics, voice and text; the application platform should refer to Mengguyun here, which encapsulates the entire three layers of things below to provide developers with computing power.

If cloud computing virtualizes computing power into a stream, and users can turn on the switch to obtain the computing stream, then for artificial intelligence, I believe the same is true, and it can also be understood as a new expansion of the cloud.

Users can use artificial intelligence by turning on the switch without considering hardware, interfaces, algorithms, just like the cloud, plug and play. I think this is a good vision.

As for what to do if you plug it in?

I think it should be up to you.

......
Chapter completed!