On July 12th, the 8th China Finance Summit was closed in Beijing by a number of financial and mass media companies.
As the most influential platform for ideological exchanges in the economic field, this summit has attracted much attention and once again became the focus of attention. The summit will be based on the theme of “New Era: Great Changes and New Dynamics”. A total of 4 main forums and 13 theme sharing and parallel dialogues were held. More than 1,000 political and business leaders, academic experts and media representatives gathered here to judge trends and explore new issues of the times.
Despite the impact of the global macro situation, the Chinese economy has risen to the challenge and continues to improve its growth. Whether it is artificial intelligence, big data or the Internet of Things, or a new technological revolution, it is able to inject traditional energy into China. The old engine with a weak economy makes it alive and well.
Only by courage to innovate and constantly change can we break through the bottleneck of economic development. Many participants at the summit shared their views on the topic of innovation and change.
AI landing, dialogue business future
1、Moderator: As a company specializing in storage main control chip, Beijing Starblaze Technology faces artificial intelligence application scenarios. What are the technical enhancements?
Daniel Sun : First of all, the first transmission speed is getting faster and faster, and the computing power is getting stronger and stronger. The second of all this data is valuable and must be stored, analyzed, and transmitted.
Starblaze Technology is a company engaged in the research and development of high-end controllers. It has seen the opportunity of security monitoring in the development of high-end controllers. Now the security monitoring still uses human observation to find the target and then track it. Now that AI is introduced, this person will be tagged, but it encounters a difficulty that may not seem easy to overcome. In the prefecture-level cities of 8 million cities, the amount of snapshots and video data generated every three months. It is 14,000 TB, and the suspects found are more than 100 K. For this application, we have the natural advantage of the storage controller. The storage controller can regard it as a small computer. It has a CPU on its own. The flash memory particles controlled by the storage controller are hard disks, which is a small host. After the data you sent, after the label is finished, I will manage the database according to the method of sorting in different ways. Each hard disk is equivalent to a small DBA administrator. As long as you insert more hard disks on your machine, send an instruction. , telling you that the required tags can help you with large-scale searches, and can also be tracked and processed with custom matching functions. At the same time, we will also make a product release at the World Artificial Intelligence Conference in August this year.
2、Moderator: The second question I would like to ask you, artificial intelligence has reached the stage of empowering business, helping them to achieve digital and intelligent, then how should artificial intelligence actually solve industrial problems?
Sun Tang: Let's first talk about it from the perspective of technicians. Artificial intelligence can layer it. Let's talk about weak artificial intelligence. The scene of landing is very common, similar to story machine, educational robot, face access control and very terminal. Small applications are already very broad. The other type is the cloud intelligence that everyone doesn't usually see. It is similar to GPU and NPU. They are usually used in the cloud to do big data processing and advertise the stereotype. This is a relatively large amount of computing power.
The second is the edge computing that 5G ushers. More than half of the calculations in the 5G era will be completed on the edge. There will be a lot of application landing opportunities at the edge. Security, fire, medical and similar logistics robots are already in this market. There is no large-scale extension. With the advancement of technology in the future, I personally feel that there is a great room for development, and we also see limitations. Now there is a joke that there is a lot of intelligence in the manual, because artificial intelligence accepts a large amount of data after training results, this data requires people to label, make advance reservations, requires a lot of manpower to spend on it, this process is currently There is no standardized process and platform for the stage, and all customized scenarios should be tuned according to the scenario. For example, if you have a very good result in a lab, you may not even have 10% in the actual scene. You need to use a lot of manpower to do the iteration of the parameters. This is the pain point of the real application. Some technical directions may be overcome in the future. This pain point is similar to the multi-target confrontation generation network, which is relatively hot in the industry.
Another deep intensive learning, I will not elaborate on these technical terms. Once a target is trained for the first time, when he deploys to the device, he can intelligently learn the extension according to the current situation. This scene encounters some pain points that can upload data and make some minor adjustments without the need for manual participation. This is the basic condition for artificial intelligence to fall.
Moderator: I can feel that the data in the information age is the new force in the information age, and can give us a lot of strong momentum. Then I would like to ask you, the domestic chip in the past some time due to certain events, everyone’s attention will be greater than before. The improvement, what do you think of the future development prospects of domestic chip companies?
Sun Tang: To be honest, before the year 2015, the domestic chip industry did not pay much attention to it. The real heat generation should be after some events. The domestically produced chips ushered in a rare opportunity. I can recall the core. Let's make an example to illustrate the history of domestic chips and possible future development prospects.
Starblaze Technology was established in November 2015. In less than three years, there have been three generations of chips in one-off production. Our chip is one of the few controllers with a consumption level of more than 7500 points. Industry customers and consumer customers have already matured their plans to develop and develop. If you have enough professional attitude and strength, you can get a certain market space and effective development. The first point we think must be professional, focus on the industry you are engaged in, and truly make this technology deep. In less than three years, I applied for more than one hundred invention patents, all of which are independent intellectual property rights. At the same time, we also see that because the integrated circuit industry chain is very long, we believe that the Chinese core should be united. Only everyone can go upstream and downstream. With the opening of the industrial chain, it is possible to overcome the short-term difficulties that have been encountered and truly achieve the goals we want to achieve.
