The most popular AI in zhangshunmao is becoming th

2022-07-31
  • Detail

Zhangshunmao: human AI is becoming a human brain extension to help break through space-time limitations

zhangshunmao: human AI is becoming a human brain extension to help break through space-time limitations

19:23 source:

original title: zhangshunmao: human AI is becoming a human brain extension to help break through space-time limitations

the Ministry of Commerce, the Ministry of science and technology, the Ministry of industry and information technology, the national development and Reform Commission, the Ministry of agriculture and rural areas, the State Intellectual Property Office, the Chinese Academy of Sciences The "21st China International High Tech fair" jointly organized by the Chinese Academy of engineering and other national ministries and commissions and the Shenzhen Municipal People's government was held in Shenzhen, China from November 13 to 17, 2019. Senior vice president of Huawei and CTO of cloud & AI products and services attended and delivered a speech at the "emerging technologies that change the world" forum

zhangshunmao said that AI is becoming an extension of human brain power, helping human beings break through the limitations of space, time and appearance. For example, a lot of work must be done on site. Now that 5g has arrived, it can not be done on site with the help of AI, and a lot of work can be done remotely. We have also seen many reports about telemedicine, remote education and remote control. For example, in some dangerous mines, where the operation is difficult, unmanned driving can be used for remote control. Breaking through the time limit, we don't understand the ancient characters, but they can be automatically translated to us through AI learning

the following is the actual record of my speech:

the topic of my speech is easy to understand in the future. AI, artificial intelligence and machine learning have more and more layers, and 152 layers are not the most. Only when the neuron network of learning is deeper, the final accuracy will be better. Whether it is autonomous driving or not, face recognition and character recognition will achieve the required results. The more layers there are, the stronger the computing power will be, so it is called "power computing the future". Why is it called shengtengwanli? Next, please listen to my report

these are our applications in real scenes. Because the robot uses AI to classify garbage, we see that the robot can automatically sort out different garbage, such as glass bottles, plastic bags and cans, through machine learning. Robots can sort out different kinds of garbage through learning. It is very difficult to classify garbage by people. Machine learning can solve problems more efficiently. This is the second case. It is also a real case of using AI to improve and improve our work. This is a real case of Nanfang power in Shenzhen. High voltage wires need to be inspected. Once there is a problem, it is very dangerous, and workers need to climb the pole to inspect. How can we replace this work with machines? We will use the UAV. After the UAV is equipped with a high-definition camera, the UAV will fly along the high-voltage wire to detect any possible problems on the high-voltage wire. The UAV is equipped with a high request camera, AI and a 3D electronic map, so that the UAV can fly safely and will not hit the electric wire. At the same time, the types of faults that UAVs can detect can be continuously increased through learning. We have separated the software from the hardware reasoning of AI. We can download the software to the UAV, so that the UAV has different skills and can see different quality vulnerabilities. The third case is the "new sound machine" in the tropical rain forest, which can detect different sounds in the forest, such as the sound of electric saws and the sound of cars passing by. These people may be illegally cutting wood in the forest. How can we detect such a large forest? It is certainly unrealistic to rely on manual work. It can also be solved through AI technology. It can even monitor the sound of different animals. Whether the high crystallization speed of normal polyamide 6 will limit the sound of its processing window, or the cry for help when it is sick or falls into a trap, it can save rare animals

through these three cases, we can see that AI has gone beyond the traditional information tools and support systems. All walks of life are doing information work, but to a large extent, information is only used as a tool or support system, not into the production system. This is because AI technology has really become our productivity, entered the field of industrial equipment, become a basic productivity, and changed the production mode of all walks of life. AI technology is a real productivity revolution for all walks of life. The first trend is that AI becomes a real productivity

the second trend is computing power, that is, whether a person is intelligent or not depends on his learning ability. AI also needs computing power to determine his learning ability. The human brain has 80billion neural networks. How much can AI's ability and chip's ability turn waste into energy? AI can not replace the human brain to a large extent. The computing power is not powerful enough, the power consumption is still relatively high, and the power is not as low as our human brain. The size of computing power determines the size of productivity. We can also see that Google has published more and more papers worldwide, which is inseparable from its strong Google cloud computing power. It can quickly calculate the results of difficult problems, especially in the case of big data, a large amount of data will appear. It requires computing power to gain insight from this large amount of data

the third trend is that computing power is not only in the data center, but also everywhere. From the cloud to the edge, to the end, and even to the end side, the number of chips used by arm is hundreds of times more than that in the center side. At present, there are more than 20 billion arm chips, including Huawei Kirin. At the same time, we see that the computing power on the end side will be more and more, and the deployment on the edge side will be more and more abundant. Cloud, edge and end computing will be more efficient if the same architecture is used. The scale on the side is smaller, faster and shorter. More complex problems will be calculated on the central side

the fourth trend is that AI is becoming an extension of human brain power, helping human beings break through the limitations of space, time and appearance. For example, a lot of work must be done on site. Now that 5g has arrived, it can not be done on site with the help of AI, and a lot of work can be done remotely. We have also seen many reports about telemedicine, remote education and remote control. For example, in some dangerous mines, where the operation is difficult, unmanned driving can be used for remote control. Breaking through the time limit, we don't understand the ancient characters, but they can be automatically translated to us through AI learning. Breaking through appearances, some superficial phenomena can gain insight into deeper problems at the bottom of the data through big data analysis

the fifth trend is that AI is a very good technology, but any technology or tool will have such problems. If it is used in inappropriate places, it will have negative effects. AI involves a large amount of data, which involves the security and privacy protection that everyone is concerned about. It is a very big threat. In this regard, how the technology deals with this problem has become more and more serious, not just to make a gesture of protecting the environment. For example, we are now used to mobile payment. How can we prove that you are you in mobile payment? It is possible that we should listen to you at that end, watch your photos through videos, and let you say a few words and move. Even so, it may be false, At that end, another person may use your voice to simulate your voice. What you see in the video is your friend's image. In fact, it is not. He may have intercepted your friend's image and moved. These are all appearances. This has become a real threat. How to solve these problems

we see the great value and role of AI. We began to study this technology many years ago. In recent years, we have continuously released new chips. The most important chip in AI is the shengteng processor. This chip provides us with powerful computing power for training, learning and reasoning. Why does it have such powerful computing power? It uses the Da Vinci architecture. Vector computing and vector computing are usually used for the calculation of neural networks. On this basis, we introduce matrix computing, which can be large or small. When the scale is large, it can be deployed in the cloud, and when the scale is small, it can be deployed to the end. Kunpeng processor is a general-purpose computing processor released this year for general computing power. This year, we also released the world's fastest AI training cluster called atlas 900. There is a standard testing machine RESNET in the world. We have reached the world's first computing power level. We can calculate the results in less than one minute. Let's see how fast it is in the case. This is the starry sky in the southern hemisphere. There are 200000 stars in the starry sky. We need to find a certain characteristic star. How to find it. In other words, with the cluster trained by Huawei atlas 900, we can make the machine complete the task in 10 seconds. Before, it took more than 160 days for half a year to find the computing power. Now it can be found in 10 seconds. This shows how powerful AI is and how fast it is

just now, we mentioned the privacy and threats of AI. We recently released a white paper, "think ahead: security and privacy protection of AI", which published Huawei's practices in security and privacy protection governance and methods to jointly manage this problem. We are also very proud to see that the Ministry of science and technology announced that Huawei will take the lead in building a new generation of national AI open innovation platform for basic software and hardware, which is a series of solutions based on the Pentium processor and Da Vinci architecture, including hardware, basic software, mindpole computing framework and developer oriented development platform, Based on this software and hardware platform, a large number of developers can easily develop their own models, applications and algorithms in their own scenarios. As a national basic open innovation platform, they have started to deploy in Pengcheng laboratory in Shenzhen. At the same time, in order to help developers from all walks of life to use and make good use of this platform, understand it, support it and get familiar with it, We also released a "The fertile soil program provides us $1.5 billion to develop 5million developers around the world, including enterprise developers, individual developers, start-up companies, and school students. They began to learn shengteng and Kunpeng technology from school. We have a real case of nttdata, a Japanese company whose wheat stems have high strength and good flexibility. It is based on Huawei The hilens machine vision service on the cloud has created a solution for a smart car 4S store. This 4S store is a digital 4S store. When a customer walks into the store, he can analyze the customer's characteristics, know who is coming, and know what kind of experience he has had in the 4S store in the past, including whether his expression is happy, unhappy, anxious and irritable, We can let the 4S store staff provide more appropriate services to customers. Let's watch a video. This is an example of many of our AI development kits. In fact, we have many such developers around the world who develop their own innovative solutions based on Huawei's platform. Next year, we will hold a global developer conference in the Convention and Exhibition Center on February 11. We will open and share more technologies for developers to use. Welcome to join us, Thank you

source: Sina Finance

Copyright © 2011 JIN SHI