The edge intelligence of Da Shengwei, the founder

2022-07-28
  • Detail

Da Shengwei, founder of Xiaoyi Technology: edge intelligence is the last kilometer of artificial intelligence

Da Shengwei, founder and CEO of Xiaoyi technology, introduced the achievements of Xiaoyi technology in the past few years at the AI world 2018 World Artificial Intelligence Summit, and took this as a starting point to deeply analyze the future development of computer vision. It also introduced the three-year high-speed development process of small ant technology to become a unicorn, deeply analyzed the future development of artificial intelligence computer vision, put forward the new concept of "edge intelligence", pointed out that edge intelligence (EI) is the last kilometer of artificial intelligence (AI), and made a wonderful speech entitled "edge intelligence (EI) - the last kilometer of artificial intelligence", which mainly includes the following points:

I. edge intelligence, Reconfiguration intelligent algorithm engine

what is edge intelligence? As the name suggests, it is the edge of interconnection and IOT. The most edge is a large number of sensors. CMOS is one of the most important sensors. We focus on the image field, so the last kilometer of computer vision application is the field that Xiaoyi technology focuses on. The development of this field must adopt the combination of edge intelligence and cloud, that is, the combination of edge and cloud, that is, the edge intelligence architecture

what are the applications that are really closely related to people's lives? A passive needle can't stop at any position. Generally speaking, what is related to people's going out is food, clothing, housing and transportation. How will AI in food, clothing, housing and transportation be related to computer vision? Clothing, food, housing and transportation are just a few common scenarios such as smart home, intelligent travel and offline retail. These scenarios all have complete processing methods such as hardware, software and computer algorithms. Xiaoyi technology reconstructs the digital space through EI, including three aspects: reconfiguring computing power, reconfiguring algorithms, and reconfiguring the business model

II. 13million users worldwide, with daily video data of more than 100pb

in the past four years, Xiaoyi technology has obtained domestic and international computer vision patents and works, and obtained a large number of overseas patents. The next development of Xiaoyi technology will further develop in the field of AI enabled video hardware, covering 168 countries and regions, and has research and development centers in China, the United States and Israel, which generate a large amount of video data every day. Xiaoyi technology is endowed with a huge amount of data. It produces more than 100pb of video data every day, and has more than 10million users worldwide. Up to now, there are 13million users worldwide, including 500000 commercial users. The main partners of Xiaoyi technology include Microsoft, Google, Alibaba, Baidu and Qualcomm

III. EI's three core competitiveness: computing power, algorithm self growth model and enabling business

first, computing power is the "core capability" of an artificial intelligence enterprise. Global computing power of cloud + end. Xiaoyi technology has sold more than 1300 devices, which will grow exponentially in the future. The computing power of these devices will also grow exponentially

secondly, the deep neural algorithm of Xiaoyi technology adopts the deep learning neural network above 1000 layers in the cloud, and the embedded and small bit neural network is more used in the terminal. Xiaoyi technology has reached the highest equivalent computing power per gflop, which should be very leading in the industry

third, business enables industrialsolution, such as travel industry and shared vehicles

IV. what can the reconstructed space do for people? Real time active decision-making

IOT has existed for some time. Xiaoyi has been committed to providing IOT hardware, software and the whole system since its inception. In fact, these early products only completed the edge perception. Small ant technology starts to process and calculate at the edge. The whole intellectualization also combines with other sensors to aggregate data, which is called edge intelligence. Suppose there are 10000 chain stores, or 1million cars in the whole region. Gather the number of these cars and the number of stores in the cloud to analyze big data and filter out noise, so as to form a better model and train better algorithms, so it is called end cloud combination. This is from edge awareness to edge computing, edge intelligence, and finally to end cloud integration

v. advantages of edge computing: new Moore's law

edge computing can be processed faster and better. In 2020, there will be 50billion IOT devices in the world, and 50% of the devices will analyze, identify and store data on their own. The proportion reached today is even far less than 50%, I am afraid it is only 10%-20%. In the next twoorthree years, the rail transit industry cluster will be formed by the strong wind of investment and development; Based on Yuanda Kejian

there are indeed some bottlenecks in the cloud, such as energy consumption and efficiency. However, the gap at the edge is still very large, and there is also great room for improvement. If it can be solved well, the pressure of the network can be greatly reduced. Suppose that there are 20 intelligent devices in a store, such as cameras or audio devices, which cannot be calculated locally, but must be transmitted to the cloud with uplink bandwidth. I believe this network will be blocked. Moore's law is the only magic weapon of small ant technology. In the next five to ten years, edge computing will directly convert carbon dioxide or methane into green substrates. A new Moore's law was born because there was no real requirement for computing power at the front end. Once there is a requirement for computing power, good algorithms and even good applications, it needs to be further improved, and Moore's law can play a role

VI. smart retail enabled by small ant edge smart retail

small ant technology provides solutions for intelligent digital space, that is, visual solutions for offline retail. These solutions include products of smart new retail stores and smart shelves, a series of complete systems related to video, and finally, on-site management and decision-making for providing software services to smart retail. Due to the development of various information technologies, especially in the early stage of the scale implementation of IOT industry, edge computing has been put on the agenda of the industry. With the application of artificial intelligence in the edge computing platform and the increasing demand for the coordinated implementation of edge computing and IOT "end management cloud", edge intelligence has become a new form of edge computing, enabling IOT applications

small ant technology is an enabler of edge intelligence. Xiaoyi provides 28 products in 6 categories under Xiaoyi brand, covering network cameras, motion cameras, VR, etc; Xiaoyi products are linked to Xiaoyi Intelligent Cloud and provide enterprise level cloud PAAS services; Xiaoyi's cloud + end basic services are combined with the industry to provide solutions for smart retail, travel, smart Park, building, home, etc

Disclaimer: the above content is transferred from other media by the website. The relevant information is only for the purpose of transmitting more information. It does not represent this view, nor does it mean that the website agrees with its view or confirms the authenticity of its content

Da Shengwei, founder and CEO of Sohu Xiaoyi technology, introduced the achievements of Xiaoyi technology in the past few years at the aiworld2018 world artificial intelligence summit, and took this as a starting point to deeply analyze the future development of computer vision

Copyright © 2011 JIN SHI