Huawei founder Ren Zhengfei recently gave an internal speech at the Huawei Noah's Ark Lab Symposium, and for the first time systematically talked about Huawei's strategy in the field of artificial intelligence.
The article mentioned that Ren Zhengfei mainly talked about four aspects, one is Huawei's artificial intelligence service direction, the second is research and development focus, the third is teamwork, and the fourth is talent acquisition.
Ren Zhengfei pointed out that we must prevent closure and must be open. Here is a US brick, and there is a European brick, a Japanese brick, and a Great Wall. No matter who the brick is, you can win it. Don't make any bricks. On this large platform of the Great Wall, allowing the river to dance, allowing the "bee" to dance, it can not subvert this platform, but activated the platform.
The following is a record of Ren Zhengfei's speech:
All of our artificial intelligence has to eat its own dog food first, and the parachute itself produces jumps first. Based on our huge network inventory, artificial intelligence is now focused on improving our services. Service is the company's largest stock business, and the most difficult business. Artificial intelligence can play a role in the service field first. Which company in the world has such a large amount of business and data to compete with us? Through the accumulation and improvement of services, it is possible to produce the world's strongest artificial intelligence experts after five years, and at the same time liberate a large number of quality service experts to invest more strategic units in the attack on Shangganling.
Therefore, artificial intelligence should aim at serving the main channel and make up its mind to spend money to build the company's inherent capabilities. First, do not do things outside the border, and do not make small products in the society.
First, the huge stock network is the best stage for artificial intelligence
Why focus on GTS and the ability to artificial intelligence in the service field? For the increasingly large and increasingly complex networks, artificial intelligence is the most important tool for us to build and manage the network. Artificial intelligence should also be focused on the main channel of the service. This development of artificial intelligence is to develop the main channel business. We want Put it to this height. If artificial intelligence supports GTS to do a good job, our own problems will be solved after five years, and our artificial intelligence is world-class.
First of all, it is the ability to solve the network maintenance, fault diagnosis and processing of our huge network inventory in the world. Our global network stocks are $1 trillion and increase by hundreds of billions a year. The capacity is getting bigger and bigger, the traffic is getting faster and faster, the technology is more and more complicated, the level of maintenance personnel is getting higher and higher, the experience requirements are getting richer, and there are no such talents, artificial intelligence, and great prospects. .
We are now using an IP network. IP is sacrificing latency to reduce costs. Routing is awkward. Where is the problem? Where is the problem? I don't know. There is a problem in the UK, probably in Germany. Virtualization software and hardware are decoupled. In the future, more attention should be paid to sub-health checks in the network. Tomorrow, the network will become more and more complicated, and it will become more and more uncertain. The fault does not know what is going on.
Huawei accounts for one-third of the global network, and it is difficult to maintain such a large inventory network. The equipment on the Internet is from old and young, to fashionable youth, and new and new human beings. If there is no self-learning of artificial intelligence, and the knowledge and skills are continuously abandoned, how can this network be maintained by people? People can't remember so many accident models. Therefore, we must build this ability. We must work on the problem of automatic diagnosis and automatic detection of hidden troubles. Otherwise, our organization will be very bloated in the future. We must dare to invest here.
By learning artificial intelligence, experts can focus on solving the most critical 10% of problems. Some simple questions can be implemented automatically, so that the service experts can focus on solving the key problems. The reduced compilation can all be used to recruit artificial intelligence research to recruit scientists and doctors (including, of course, Dr. Tea, pre-doctoral).
Our current base station installation is the installation of the field hardware, we do the total adjustment in Xi'an, Romania... If you find problems in the future, you don't have to go to Romania. The problems that occur in our local area, after self-learning of data from all over the world, the system can adjust and solve the problem itself, and then report the results. Through expert analysis and training, we correct the structure of machine algorithms, and improve algorithms in dealing with problems. The most important thing is to let the machine have the ability to learn, not just the ability of people to learn.
It is very fragile to repair a house with cement alone. The wind can blow down. Adding sand to the cement and adding some stones is very hard. You have to do concrete works. To do artificial intelligence, we must be close to the actual, close to the needs, close to the customer. Noah's Ark Lab should have a team in each GTAC, and work with the service experts every day to do troubleshooting, to understand what is the fault, how to eliminate the fault, how their data model is, they do not Will tell you a story after solving the fault. In addition, you should be familiar with what the network is. If you are not a network expert, how can you find faults through artificial intelligence? It’s awkward to know the problem. The way to solve the problem is embarrassing. This is not the GTS landing, but the responsibility of the 2012 lab. The corresponding team of GTS can invest and cooperate.
Second, artificial intelligence in the network large traffic forecast makes network planning and optimization from passive to active.
Taking Chengdu as an example, with the development of video services, 4G users increased their traffic by 75% in the past year by 70%, and the average download rate of users in Chengdu increased from 35Mbps to 40Mbps. So, the customer's challenge is how to protect and enhance the end user experience while the network traffic is growing rapidly. As the services carried by the network become more and more dynamic and more dynamic, it is necessary to use artificial intelligence to actively predict, to actively discover the traffic hotspots in the next few months and make adjustments to the network beforehand.
In the future, the network is a data center-centric network. In the planning and design of the network, the network has a long delay in sharing, and when it is closer, the delay is less, but the data center is more, and tens of thousands of data centers. The transfer of data between them is a complex algorithm problem, which also requires artificial intelligence to play a role in network planning. Why have I been encouraged over the years? I have to learn from aerospace, geography, mapping, biology, etc. to enter the service system, that is, to dare to solve problems with the most advanced tools and methods. By using advanced tools, the topology map of the network is taken out, the satellite map is taken out, and artificial traffic is used to predict the output of a large flow rate. Then, the advanced case of Korea and the advanced case map of Sichuan are taken over. , you can predict where the network's traffic opportunities are. The accidental seedlings were discovered through abnormal flow changes.
Now our network optimization model is after the fact. According to your traffic prediction and automatic planning examples, you can predict in advance, so that network adjustments can be made before user congestion occurs, and problems can be avoided in advance. I know that you can't finish it in one step, but we will go forward step by step, we will definitely find the window of opportunity. Although some of the content is just that the demonstration has not entered a comprehensive and practical state, I believe that today's vacation is the truth of tomorrow. I support you. We must find the most practical and simple way to serve the world. We have built this big mechanism and team, and with such a large capacity, we can better serve our customers.
Therefore, the application of artificial intelligence in the service is to analyze the network fault diagnosis, the second is the guidance to the network planning network, and then the translation of the technical data. Our artificial intelligence should be done first, and we will experiment with our internal business one by one. This year, this piece will be made a little. The next year, the piece will be made a little. The more difficult it is, the more difficult it is to do. I will give the budget to the inside and make up my mind. If money builds these skills in service, there will be a future. Offense is the best defense. When we attack in this way, the threshold is too high for other companies to keep up. After we have more than 150 billion US dollars in 2020, we will become a slow cow, and we will not grow so fast. At this time, if artificial intelligence is used well, we will control the number of people and increase efficiency and efficiency. Then our company is still a good business. situation. At this time, the team we cultivate can be killed and put more strategic troops into the new Shangganling.
Jiangmen soundrace electronics and technology co.,ltd. , https://www.soundracegroup.com