CHINESE ENGLISH
Industry News

Location:Home > News > Industry News... > AI chips into the in...

AI chips into the industry is new air porthow should Chinese enterprises layout

Source:Updated:2017-11-22 08:51:25

AI chips are becoming a new wind in the industry. On November 15th qualcomm announced that it would invest nine Chinese artificial intelligence algorithms + chip innovative companies. Previously, the Chinese company hangzhou state announced the launch of the voice AI chip, and China's Cambrian company also released three AI processor IP and an AI software platform. The boom has intensified abroad. Not long ago, Microsoft announced the development of AI chips, which will be first applied to its new generation of Hololens AR devices. Intel, nvidia, qualcomm, Google, apple and other giants have also joined the AI chip fight. How will AI chip pattern evolve in the future? How can Chinese companies share a slice of the pie? These are the focal points of the innings.
AI chips into the industry's new air port, how should Chinese enterprises layout
Application drives special AI chip development
 
"AI is one of the main development directions for new technologies and applications in the field of information, and the IT giants are laying out AI chips to seize control of the supply chain of the AI era core components." Liu xinliang, senior analyst at cedi think tank integrated circuit research institute, told the China electronic news.
 
Today's AI chips are very different from those used in the past. Some analysts believe that with the traditional calculation through fixed process solve the problem of "ok", deep learning algorithm to simulate the human brain to solve those who depend on the probability of "uncertainty", and the computing power needed for the problem solving process, but do not agree with traditional chip do well. That is why traditional chip giants such as Intel, nvidia and IBM are developing new dedicated AI chips.
 
Intel China research institute, said dean Song Jijiang AI now required computing power, some companies are adopting the graphics accelerator to do training, identification, one of the difficulties is as computing power has increased, power consumption is bigger. While many smart devices are mobile, whether it's a driverless car or a service robot, it requires a lot of time. And do AI intelligent device has three steps, recognition, reasoning, action, so it is not only a need for identification, analysis, mining information, computing power and the contradiction between power consumption becomes more and more highlighted.
 
"How to solve the contradiction between computing power and power consumption? One is the use of software, while artificial intelligence is a soft and hard combined system, which is not pure software. Software can be optimized for algorithm and cut down the computation. The other is the way hardware is used. Hardware also has two ways, one is in the same way that a general CPU, the way gm but power consumption is very big, the second is to use a dedicated ASIC chip, it can achieve very low power consumption, performance is very strong." "Song said.
 
Based on this, we can understand why these apple, Google, and Microsoft are working on artificial intelligence chips for their smartphones, self-driving, AR/VR. To create a better user experience and better performance, more and more IT giants have chosen to do everything from chips to software to hardware and software, in order to create the "iPhone effect" of the AI era.
 
In fact, in the era of multiple intelligent devices, various dedicated AI chips need to be enriched and diversified. According to Tirias Research, almost all electronic devices used by people will need artificial intelligence support by 2025. This means that a variety of vertical scenarios in multiple terminal markets, including mobile phones, smart homes, security, autonomous driving, and other areas, have plenty of opportunities for AI chips. On Nov. 15, qualcomm announced a venture capital investment in nine Chinese artificial intelligence algorithms and chip innovation companies to preempt the entrance to the various scenes in the artificial intelligence ecosystem.
 
Liu said that AI chips are showing such a development trend, one is in the special AI chip aspect, will present the application driven characteristics. In terms of application scenarios, autonomous driving is one of the most promising applications, both from market scale and technical maturity. According to the data, automobile industry accounts for about 1/6 of the total GDP, while autonomous driving is the focus of the automobile manufacturers' scrambling to lay out the layout. Autonomous driving will be the biggest air outlet in the recent AI market. Second, in terms of general chip, will present a CPU, GPU, and even the FPGA integration, the development trend of general-purpose CPU will increasingly prominent graphic processing, speech recognition and deep learning AI computing power. AI now is still in its early development stage, UBS (UBS) predicts that in very large server, self-driving cars, robots, unmanned aerial vehicles, smart phones and the Internet of things (IoT) device leads, the AI chip the size of the market in 2021 will be from $6 billion to $35 billion in 2016.
 
The market that looks like early spring, mordaun early, more early pedestrian. Investment in the AI sector is estimated to be between 26 billion and 39 billion in 2016, three times more than three years ago, according to a June forecast by the McKinsey global institute.
 
Chinese companies should accelerate AI chip layout
 
China should accelerate AI chip layout. It should be said that the advent of the era of AI, to the chip industry in China has brought the new change, whether to apply for driver chip dedicated AI, or is exhibiting the CPU and GPU fusion general AI chip area, has a chance to China.
 
In the field of universal chips, China has been hoping for a bigger breakthrough, and it should be said that the advent of the era of artificial intelligence has made it possible for the dream to be a breakthrough.
 
Liu Xinliang talked about the changes of the industry over the years, moving from the PC era to era and then to today's Internet age, every industry change created many new opportunities, also a new chip giant achievement. The PC era, for example, is the way for chip companies such as Intel and AMD. In the mobile era, the rising opportunities flow to the advanced qualcomm; In the era of artificial intelligence and Internet of things, there are no dominant giants in the world, so in this period, Chinese enterprises have the opportunity. With such as huawei, spreadtrum communications manufacturers, has experienced the era of mobile Internet technology accumulation, team, talent, capital and ecological and market advantage, choose to cooperate with professional AI technology company, the layout and the preemption is not yet mature market of AI. For example, huawei's cooperation with the Cambrian is likely to make more breakthroughs in the general field, and the newly released unicorn 970 from huawei has joined the NPU, an artificial intelligence unit from Cambrian. In the first three quarters of this year, huawei's mobile phone shipments were the second highest in the world, with 34.2 million, 39.1 million and 38.5 million respectively in the first, second and third quarters.
 
China also has plenty of opportunities in the area of dedicated AI chips, especially in the application-driven technology market, where Chinese companies have been adept at cutting through applications. Director of Intel's global vice President and general manager of China Yang xu in an interview with "China electronics news" reporter said, artificial intelligence to deal with all sorts of massive amounts of data, in the face of different application scenarios, in the face of so much data so quickly change, the market, the Chinese company's high-speed operation, low cost and large scale did not exist in other markets.
 
October 31, the hangzhou core chip GX8010 released its first speech in shenzhen AI, the chip designed for Internet applications, carrying NPU processor (neural network), DSP (digital signal processing (DSP) technology, with low power consumption, can be offline, mobile advantage. The chip is aimed at smart speakers, smart TVS, smart toys/preschool.
 
A few days later, another highly focused AI chip company, Cambrian, released a smart solution for terminal applications and cloud-based solutions. Included in the terminal part of the low power scene visual application oriented Cambrian 1 h8, has wider and higher performance of commonality "Cambrian 1 h16," and for "Cambrian 1 m" in the field of intelligent driving. In the cloud, the Cambrian also released the high-performance machine learning processor chip "Cambrian MLU100" and "Cambrian MLU200". Goal is to strive to Cambrian CEO Chen Tianshi revealed that the Cambrian in 3 years occupies 30% share of China's high-performance smart chip market, and make the world more than 1 billion sets of intelligent terminal equipment integrated with the Cambrian terminal intelligent processor.
 
It is important for Chinese chips to make big breakthroughs. Wu gansha, a former dean of Intel's China research institute and CEO of the command and technology company, said that the success of a chip includes both the success of the technology and the success of the ecosystem. Therefore, China's AI chip must be able to achieve success in the market.
 
Huang zhijie, CEO of hangzhou national core, said that the AI chips should actually be landed, and that the NPU is not enough. Because the AI interaction is a very complicated process, in addition to the neural network computing access include sensor, signal processing, detection identification, as well as the decisions and feedback on the level of software, many links, each place isn't need algorithm and computing characteristics, puts forward the core "comprehensive integration, the whole stack through" strategy.
 
The Cambrian hopes to speed up its cloud-based interactive AI landing in a technological and ecological way. "It's hard to be successful in the market without matching apps and software." "Chen said. The current Cambrian partners, including alibaba, huawei, lenovo, ARM and other Internet and hardware companies, also include the science and technology of flight, absenteeism, commercial soup and so on.
 
According to shen nanpeng, an investor, there are already 1,000 AI start-ups in China, with 50 per cent of all investment activities in China linked to AI. There is a surging AI era in the world. China has talent, money and a large number of innovation teams. We have every reason to believe that China will have great opportunities in the future.
Product|Technical|News|Propaganda|Partner|About

Copyright © 2014 Quanstar Intelligent Controls(shanghai)Co.LTD All rights reserved ICP: Shanghai ICP No. 10214886

Address: 152 Ring Road, Shanghai Comprehensive Industrial Development Zone

Tel: 021-57472600    E-mail:Mark.liang@quanstar.cn

Website buildingTrueland