![<?echo $_SERVER['SERVER_NAME'];?>](/template/twentyseventeen/skin/images/header.jpg)
Professor Wei Shaojun recently said when talking about the status quo of China's AI (artificial intelligence) chip industry: "Now the hype of the AI ​​chip is overdone. Today, even some of the AI ​​chip founders will become the 'martyrs' of technological change." Wei Shaojun’s words are not groundless. In just one or two years, the domestic AI chip head startups’ financing quota exceeded RMB 2 billion. At least 45 startup companies worldwide have developed AI chips (voice interaction and autopilot), and five companies have raised more than US$100 million. According to estimates by CITIC Securities, the AI ​​chip market will reach US$14.616 billion in 2020. Artificial intelligence technology is rapidly emerging with AI chips as its carrier.
01
AI chip ecosystem
In a broad sense, chips that can drive AI programs are called AI chips. In a narrow sense, the AI ​​chip is a specially designed chip adapted to the AI ​​algorithm.
From the application level, AI chips are mainly divided into cloud AI chips and end AI chips. Cloud AI chips are used in cloud servers and data centers; AI chips are used in smart devices and IoT devices. In the future, artificial intelligence will be greatly popularized in our daily lives. As Nvidia founder Huang Renxun said: "In the future, AI and AI chips will be everywhere: coffee machines, mugs, microphones, even earrings, shoes, etc. Small objects will be smart."
The cloud AI chip is characterized by its powerful performance and can simultaneously support a large number of operations, support for image recognition, and voice and video processing. The AI ​​chip needs to be embedded inside the device so that the device can have AI capabilities without networking. The significance of the AI ​​chip in artificial intelligence can be understood as that the engine is in the car. Artificial intelligence theory has been proposed for many years, because it requires a powerful "engine" drive, so many years did not really run up until the emergence of AI algorithms, big data and AI chips.
The break of artificial intelligence was in the 2012 Olympic Games ImageNet Challenge of Computer Vision. Professor Geoffrey Hinton from the University of Toronto and his team used GPU chips and deep learning algorithms for the first time to become an AI. An important node.
In the 2015 ImageNet competition, the Microsoft Research Asia team used GPUs and deep learning algorithms for the first time to enable computer image recognition over humans. The error rate of human mapping is about 4%, and the error rate of the champion team's machine recognition is 3.57%. After the rise of image recognition, a series of AI applications such as video recognition, speech recognition, translation, and voice assistant emerged.
The outbreak of AI chips will affect at least four application scenarios: home/consumer electronics, security surveillance, autopilot, and cloud computing.
Figure I. AI chip industry ecosystem
Mobile communications
Huawei Unicorn 970, released in September 2017, is equipped with an NPU (neural network processor) AI processing module and is the world’s first mobile AI chip. NPU is 25 times faster than CPU, and energy consumption is increased by 50 times. Apple released iPhone X with A11 processor in 2 weeks, the first AI chip with face recognition. ARM immediately followed by the introduction of two AI chips for mobile-end object detection and machine learning processors. The impact of ARM's action is enormous because more than 90% of the world's mobile phone chips use the ARM architecture, including the Unicorn 970 and the Apple A11.
Followers are MediaTek and Qualcomm. The HelioP60 from MediaTek supports AI's computer vision and face recognition. Qualcomm's A1 engine based on the Opteron chip packs all the SOC hardware and software in a mobile phone. The AI ​​chip has become an important spoiler for mobile phones on the Red Sea battlefield.
Security Monitoring
The security market exceeded 635 billion renminbi in 2017, a year-on-year increase of 17.6%. Jiadu technology "artificial intelligence technology white paper" pointed out that in 2017 the domestic high-definition camera shipped 100 million.
Because AI can quickly structure video, almost all AI chip startups now use security as one of the core application scenarios and have launched AI chips embedded in security surveillance cameras. The security giants Hikvision, UOB, and Yushi Technology are not only partners of many AI chip companies, but they are also pushing the pace of security and AI.
Autopilot
The AI ​​chip is becoming the core of the automated driving computing platform. In this area there are currently three major AI chip forces: Nvidia, Intel and Horizon. Nvidia and Intel are familiar with each other. Horizon is a Chinese startup company and former chief executive of Baidu Research Institute Yukai is the CEO. Horizon's Hugo Autopilot platform used Intel FPGA processors early on and is now building its own BUP architecture and introducing the "Journey" of the AI ​​chip that follows this architecture.
cloud computing
AI chips can provide power for the Internet, such as online translation, match verification, image search, and so on. Behind the AI ​​chips are the support of AI chips. Intel, who missed opportunities in the mobile space, did not hesitate to spend heavily: $16 billion in acquisition of Altera (FPGA), $400 million acquisition of Nervana (neural network processor), and subsequent acquisition of Movidius (visual processor, acquisition of funds is unknown). Both AI chips and AI chips are involved. Others, such as Qualcomm, ARM, and MediaTek, have also entered the field one after another.
Figure 2. Two major application areas of AI chips
02
Domestic AI chip major players
Cambrian: A start-up company in the background of the Chinese Academy of Sciences, the founders Chen Tianjun and Chen Tianshi are brothers, all of whom are doctors in the Institute of Computing at the Chinese Academy of Sciences. Established in Beijing in 2016. The Camry 970 chip is equipped with the Cambrian AI chip. The Chinese Academy of Sciences also sent a congratulatory message to Huawei. The two brothers graduated from the China University of Science and Technology class. Tianshi studied artificial intelligence algorithms and Tianhe learned to do computer chips. A piece of hardware, a piece of software, and the development of artificial intelligence chips. The name "Cambrian" is the metaphor of the future of artificial intelligence in the era of the explosion of life in geology.
Shen Jian Technology: Shen Jian Technology, the first camp of China's AI chip, is a start-up company with Tsinghua background and FPGA technology. Established in Beijing in 2016. Founder Yao Song, the 90th Tsinghua University hegemony, gave up the invitation of the American elite Carnegie Mellon University to pursue a doctoral degree and chose to start a business. In 2017, it obtained A round of US$10 million in financing, dedicated to the research and development of deep learning processor and compiler original technology. There have been tens of millions of orders, and two AI chips are in volume production.
Horizon Robot: The founder is Yu Kai, former vice president of the Baidu Deep Learning Institute. He once led Baidu's speech recognition, image search, Baidu brain, Baidu unmanned and other projects. The products include hardware, software and algorithms. The team includes former Nokia Mobile Phone Development VP, founding member of Facebook Artificial Intelligence Research Lab, Baidu Chief Architect, and more than 20 experts in deep learning.
Bit Continent and Kennan Yuchi: Both are producers of “mining machine chipsâ€, which are now very popular online virtual currencies. Bitland was established in 2013 and its total revenue rose to 2.5 billion U.S. dollars in 2017, accounting for more than 70% of the mining chip market. In 2017, the first 28nm process AI chip "SOPHON" was introduced. Jinan Minzhi is the world's second largest bitcoin mining machine manufacturer. In December 2017, it pre-released the AI ​​chip KPU (Knowledge Processing Unit) as a company that manufactures miner chips (self-proclaimed blockchain dedicated chips) and mining machines. , Jinan Minzhi accumulated nearly 3 billion yuan in financing. Currently has applied for the listing of the new three boards.
In addition to the above start-up companies, there are also some traditional chip manufacturers that have joined the AI ​​chip team. Hass, Hangzhou Guoxin, and China Star Microelectronics have all been involved, as well as Yushi Technology, Zhongtian Micro, and Guoke Micro, which specialize in security AI chips.
Table I, Domestic AI Chip Company
03
When will the boots fall?
In this great moment of "No chip, no AI, no terminal, no AI, no industry, no AI," some people remain sober. We need to think a bit more, and what we can do to prevent Professor Wei Shaojun from becoming a wake-up call, and to prevent “some or even most of the AI ​​chip founders of today will become the “martyrs of technological changeâ€.â€
Think one: where is the killer application?
Professor Wei Shaojun said, “Although AI applications currently cover production areas and various application levels of life, what exactly is the killer application of AI and the daily AI applications that are indispensable to people's lives? Until now, there is no answer.†In addition to the home/consumer electronics mentioned above, although close to us, there is still a long way to go for autopilot, and the entry of chips into automotive electronics is also a big threshold. Before the AI ​​chip, automotive electronics did not see any Who effectively broke the threshold. Security and cloud computing are more about industrial products, and AI chips embedded in security surveillance cameras are still in development. This is "the AI ​​product is much more to say, landing less", because in addition to cell phone chips, the market has not sold more than a million AI chip.
Thinking II: Will Universal AI Chips Appear?
The AI ​​solution architectures used by various companies are not compatible with each other. There is no standard AI computing interface supported. Each manufacturer's end equipment is also varied. Will there be a general-purpose processor like the CPU of that year to dominate the AI ​​chip world? Recently ARM announced that it wants to enter the field of AI chips and has an idea of ​​an integrated artificial intelligence ecosystem. At a minimum, the driving force of general-purpose AI processors comes from the following two aspects:
The cost of developing AI chip hardware architecture is very high, not all manufacturers can afford it;
The scalability of IP, the breadth of architecture support, and the standard AI computing interface are very important for the popularization of AI chips, and the development of related algorithms is not achievable overnight.
Regardless of whether ARM can make its dreams come true, at least the idea of ​​ARM is still very attractive to the ecology of AI chips:
AI hardware processing capabilities can be achieved with (or close to) the existing framework;
No additional expenses and additional licensing costs are required;
No need to change the design pattern of existing chips;
Provide more options for the industry;
Other customers using their own AI computing architecture can also benefit.
Another possible winner is Nvidia. The AI ​​chip is currently the most widely used GPU, because the GPU is suitable for single-instruction, multi-data processing, especially for AI deep learning training and reasoning. Simon See, Asia-Pacific Chief Technology Officer of Nvidia Technology Center, proudly stated: “GM is our advantage.†Nvidia’s Xavier computing platform is being adopted by more than 20 autopilot start-up companies, as well as Bosch, ZF and other supplier giants.
04
Future spoilers
Whether the AI ​​chip can become the future spoiler of the semiconductor industry, the conclusion is affirmative. Compared to Moore's Law chips iterating every 18-24 months, the AI ​​chip is evolving at a rate of nine iterations. Tang Weiwei, director of Product Strategy at Bitland, said, “Compared to the traditional chip iteration speed, AI algorithm iteration is faster. We can quickly write new algorithms to the chip for the needs of the latest algorithms and the commonality of neural networks.â€
Among the spoilers, there are large companies such as ARM and Nvidia, as well as newcomers such as the Cambrian and Shamtech. In the end, it is ARM's AI chip that can control the artificial intelligence of the rivers and lakes, or whether the upstart's flourishing flowers can express the artificial intelligence's magical power. It also needs time to observe.
ZGAR Aurora 1800 Puffs
ZGAR electronic cigarette uses high-tech R&D, food grade disposable pod device and high-quality raw material. All package designs are Original IP. Our designer team is from Hong Kong. We have very high requirements for product quality, flavors taste and packaging design. The E-liquid is imported, materials are food grade, and assembly plant is medical-grade dust-free workshops.
Our products include disposable e-cigarettes, rechargeable e-cigarettes, rechargreable disposable vape pen, and various of flavors of cigarette cartridges. From 600puffs to 5000puffs, ZGAR bar Disposable offer high-tech R&D, E-cigarette improves battery capacity, We offer various of flavors and support customization. And printing designs can be customized. We have our own professional team and competitive quotations for any OEM or ODM works.
We supply OEM rechargeable disposable vape pen,OEM disposable electronic cigarette,ODM disposable vape pen,ODM disposable electronic cigarette,OEM/ODM vape pen e-cigarette,OEM/ODM atomizer device.
Aurora 1800 Puffs,Pod System Vape,Pos Systems Touch Screen,Empty Disposable Vape Pod System,1800Puffs Pod Vape System
ZGAR INTERNATIONAL(HK)CO., LIMITED , https://www.zgarecigarette.com