Set micro-grid launch micro-channel IC WeChat public number: 'Daily IC', real-time release of major news, every day IC, every day set of micro-network, micro-into a!
1. Inventory of Chinese forces in artificial intelligence chips;
Set micro-news, artificial intelligence, especially deep learning, the explosive development in recent years, largely due to the accumulation of chip technology for many years.If not the chip technology has grown to a certain height, to large-scale Machine learning provides enough processing power that AlphaGo, which has not overcome top human players, has been dominated by technology-driven applications of communications chips, multimedia and smartphones for more than a decade. As these applications slowed down, advances in chip technology Has gradually shifted to the field of AI, AI driver effect will be more obvious chip technology.
With the advent of the artificial intelligence era, everyone has seen the future of artificial intelligence and its potential explosiveness, and has been a year of full deployment by chipmakers and Internet giants in the chip space since 2016. In the highly competitive Global AI Study Field, China is growing and artificial intelligence is rapidly becoming the hottest area for startups in China, with investors investing a total of $ 4.5 billion in more than 200 AI companies from 2012 to the third quarter of 2017. Then, at In the face of this unprecedented wave of unprecedented, which companies in the Chinese chip force can stand out and become a new generation of beach-goers?
Cambrian
The Cambrian discipline led by the founder Professor Chen Tianshi CAS team was established in 2016, is committed to building all kinds of smart cloud server, smart terminal and smart robot core processor chip company to create a deep learning processor direction, developed the first international Deep learning special processor chip, significantly enhance the performance of artificial intelligence field.Compared with the common processor used by Google, Cambrian chips for deep learning technology.Modeling experiments show that the 'Cambrian' relative to the traditional implementation of x86 Instruction set chip, there are two orders of magnitude (several hundred times) performance improvement.
August, Cambrian completed 100 million US dollars of A round of financing, becoming the world's first unicorn AI chip field start-ups.
Currently, the Cambrian series has three prototype processor structures: the Cambrian 1, the English name DianNao, a neural network-oriented prototype processor architecture; Cambrian 2, the English name DaDianNao, for large-scale neural network; Cambrian 3, the English name PuDianNao, for a variety of machine learning algorithms.With Google Artificial Intelligence benchmarking, the Cambrian goal is to make less than 1 watt power consumption of the camera, cell phone, and even watches can be the same as AlphaGo ' Smart ", and hope to have better performance, more powerful computing power, and lower power consumption.At present Cambrian chip industrialization is imminent, is expected to market in 2018, the market performance is worth the wait.Huawei and Chinese Academy of Sciences has been calculated Maintained a very close relationship, set up a 'CAS Institute of Computing - Huawei joint lab', Huawei unicorn 970 using Cambrian technology.
November 6, Cambrian Technology released three intelligent processor IP products, respectively, for low-power scene vision applications Cambrian 1H8, high-performance and has a wide range of universal Cambus 1H16 and for the terminal Cambrian 1M for Artificial Intelligence Products Meanwhile, Cambrian also released CambriconNeuWare, an artificial intelligence software platform for developers.
Star Microelectronics
Speaking of the domestic artificial intelligence chip, I must mention that this June 20 this year, the first to introduce China's first embedded neural network processor (NPU) chip star micro, which is the world's first deep learning artificial intelligence embedded video Acquisition of compression-coded system-on-chip, and named 'Starlight Smart One.' This depth-based learning chip used in face recognition, up to 98% accuracy, more than the human eye recognition rate. 'Starlight Smart One "took three years to complete.Versive Star Group gathered in Beijing, Guangdong, Tianjin, Shanxi, Jiangsu, Qingdao, Silicon Valley's research and development efforts, the use of advanced multi-billion-class integrated circuit design technology and sub-micron chip manufacturing process , TSMC successful mass production.
The NPU uses a 'data-driven' parallel computing architecture. The single-NPU (28nm) consumes only 400mW of power, dramatically improving the computing power-to-power ratio and can be used in HD video surveillance, smart driver assistance, Human-machine, robotics and other embedded machine vision field.At present, the shipments of 'Xinguang Smart One' are mainly concentrated in the field of security camera, which includes licensing some other security camera manufacturers.The future will be mainly to the car camera, drone aerial photography, Robotics and industrial cameras for promotion and application.
Horizon robot
Horizon Robotics, founded by Yu Kai, former head of Baidu's Deep Learning Research Institute, is committed to building artificial intelligence brain-based platforms - including software and silicon - based on deep neural networks that enable low-power, localized solutions Environmental perception, human-computer interaction, decision-making control, etc. Among them, the software, the horizon made a set of neural network based OS, have developed for the autopilot 'Hugo' platform and smart home 'Andersen' platform , And gradually began to fall.Hardware, the future horizon robot will also design a chip for this platform - NPU (Neural Processing Unit), support their own OS, then the performance will increase by 2-3 orders of magnitude (100-1000 times ).
Horizon robots have received undisclosed seed funds from investors including Sequoia and the legendary venture capitalist Yuri Milner. On July 1, 2016, Horizon robots received a new round of financing, which will be used to boost R & D investment in autopilot and smart home to speed up product development and landing, and to promote the development of artificial intelligence chips and systems. They are now working on establishing a one-stop artificial intelligence solution that defines 'Everything Smart' to make life easier More fun and safer.
The end product of the horizon is not just a chip, but a core control module that has the function of sensing, recognizing, understanding, controlling, making the control module a product to be integrated with the partner, in other words, Artificial intelligence algorithm solution.Business model, the current will take the product, technology and service approach, the first product to do, get first-hand user experience, and then to a broader ecological future. Instead of simply selling an algorithmic SDK or simply selling the chips, a technology platform was created to help users make the product. On December 20, Horizon robots aimed at autonomous driving and embedded intelligence Visual Processing introduced two new processors, Journey and Rising Sun, and proposed the Hugo 3.0 platform, the main service vehicle and the upcoming security and commercial market. The solution to the Horizon Robot is mainly to solve CNN, DNN in AI calculation and The RNN algorithm accelerates, providing the computational power required by the mainstream neural network structure. Its main architecture Based on proprietary algorithm developed by the BPU (brain Processing Unit).
Bit mainland
Established in 2013, BITMAIN is a company focused on ultra-high performance computing. The company successfully developed and mass-produced a variety of ASIC customization chips and complete systems. The mining machine that digs bitcoin is their An important product.According to reports, 80% of the world or even 90% of their mining machines are provided by them.In the past few years, they also have been promoting mining machine ASIC chip iteration, to meet the updated needs.Official information shows that the bit big The mainland's fifth-generation chip BM1387, which is self-developed by Lu, is the world's lowest power and highest performing computing acceleration chip with a production scale of several billion.
In November, the company released SOPHON, an artificial intelligence brand, which brought the world's first tensor acceleration computing chip, the BM1680, to enter the field of artificial intelligence , Is an important step for the bit continent.From the research and development of digital currency chip, mining machine to set up a mining pool, independent research and development of bitter mainland, independent layout and improve the industrial chain.Now, it is the best time for bit mainland to enter the artificial intelligence industry, committed to Powerful chip engineering, rapid iteration and system design and manufacturing capabilities to provide the most cost-effective, performance-per-watt AI computational power while being committed to tailoring and optimizing the full stack hardware and system solutions for the industry, resulting in significant industry reductions + AI difficulty, promote AI popularity.
Deep Kam Technology
Deep Kam Technology was founded by the Tsinghua team, the product called the 'deep learning unit' (Deep Processing Unit, DPU), the goal is to ASIC-level power consumption, to achieve better than GPU performance, the first batch of products based on FPGA platform.
From the official data provided, the embedded end of the product performance over Nvidia TK1 at the same time, the power consumption, the price is only about 1 / 4. The server-side products, the performance close to the Nvidia K40 GPU, but the power consumption is only 35 Watts or so, priced below $ 300, less than the latter one-tenth.
Deep Kam technology to create DPU-based end-to-end deep learning hardware solutions, in addition to the hardware module (custom PCB board) on the DPU chip architecture, but also built for the DPU compression architecture for the compiler chain SDK.
The application of the product is divided into two types of terminal and cloud.Which embedded products will be mainly used in UAV, security monitoring, robotics, AR and other fields, has established cooperation with a well-known UAV manufacturers.Server-side products Will be mainly for large Internet companies voice processing, image processing, etc. Now also with the well-known Internet companies to cooperate with each other in the existing room compatible with the case, the power consumption reduced by 80%, speech recognition accuracy increased by 5% -7%.
In May of this year, Shen Jian Technology completed a round of funding of A round of tens of millions of dollars. In October, it ushered in about 40 million U.S. dollars of A + round of financing. It was led by Ant Financial and Samsung Ventures. Since its establishment in March 2016, Shenjian Technology has obtained 3 rounds of financing with investors Jinsha River Ventures, Gao Rong Capital, Ant Financial Services, Samsung Ventures, Xilinx, MediaTek, China Creative Capital and others.
Che Sai technology
Founded in San Jose, Silicon Valley, USA in 2013, the company is headquartered in Shanghai and relocated to Shanghai in 2014. The main products of the company are smart driving laser radar, laser natural gas detection system, etc. The company brings together cutting-edge technologies in optics, precision machinery, electronics, communications and artificial intelligence Top expert, co-founder and CEO of the industry Li Yifan is a World Economic Forum Global Outstanding Young Persons, a Thousand Talents Program Specialist in Shanghai, 30 entrepreneurs under the age of 30. Entrepreneurs under the age of 30. Tsinghua University, Department of Precision Instrumentation, Ph.D., University of Illinois at Urbana- Research topics for the intelligent control of walking robots.Former chief engineer of the Western Digital Group, the United States Parker Group Strategic Investment Department, Lam Research robot department.
In May of this year, the company won a round of financing of RMB110 million, led by Pagoda Investment and led by Venture Partners, Pan Valley Ventures and Yuanchang Capital.
LiDAR, the sensor for laser 3D scanning in the autopilot and robotics industry, is also one of the most important components in unmanned vehicles. In October 2016, for the first time in the domestic market, the company celebrated the launch of the 32-line laser radar for autonomous driving , The performance indicators reached the international advanced level.After more than a year of development, a large number of internal tests and early customer joint testing, in April 2017, Woossei announced the 40-line hybrid solid-state product - the Pandar 40. It has ultra-small Volume, world-class measurement distance and measurement accuracy, and scanline optimization optimized for autopilot, creating a minimum vertical-angle resolution of 0.33 ° for all Lidar on the market-equivalent to a traditional 83-line radar.
In December, ChestSoft released its own solid-state lidar, the Pandar GT, and announced the completion of prototype development. He is also collaborating with and developing tests with a number of international OEMs and Tier 1. It is estimated that in the first quarter of 2018, Mass production partners officially provide A sample.
Long code technology
YARON Technology is an AI company that focuses on deep learning and builds a customized computer vision technology platform. YARON's core team is from Microsoft Research Asia, Tsinghua University, Chinese Academy of Sciences and so on, achieving major technological breakthroughs in deep learning and computer vision , The first in the world to achieve artificial intelligence for the typical flexible goods recognition accuracy beyond manpower, and created ProductAI artificial intelligence product identification platform for enterprises to provide the world's first product identification services, self-customized to map search engine search engine, and images Recognition Technical Support Currently ProductAI has successfully exported solutions in the fields of image copyright, fashion, textiles, furniture, home textiles, travel, publishing, medicine, smart filters and other languages in China, Britain, Japan and etc. At this year's WebVision International competition, one of the top events in computer vision recognition, the code-based Longines from China won the competition world championship with absolute superiority.
Features: First, a world-class R & D team composed of experienced computer scientists from Tsinghua University, technology enthusiasts from Tsinghua University and think tanks from overseas scholars focus on R & D and dozens of deep research in such fields as visual computing, machine learning and big data Achievements and cutting-edge technology patents, and actively promote the international academic exchanges and cooperation in a wide range of depth and achieve breakthrough technological innovation into massive GPU computing power to build AI super brain, with 10 million data to create the top algorithm model and build advanced data Tagging system.Secondly, based on the cloud-based scalable intelligent algorithm service, developers and data analysts can directly use artificial intelligence algorithms and docking APIs without managing the infrastructure and writing code.Thirdly, ProductAI platform can identify over 20,000 Object classification, part of the function recognition accuracy than professionals 8% higher; each GPU efficiency equivalent to 200-400 trained professionals Fourth, the platform is simple and zero threshold, free trial without burden, to the effect of satisfaction Paid monthly API usage.
In December, YSL won the 220 million yuan B round financing from internationally renowned SBC VC, the first startup that SoftBank invested in China's AI sector.
Wei Jing smart
Beijing Wei Jing Intelligent Technology Co., Ltd. (English name Vizum) was established in June 2016, is a commitment to the study of intelligent visual perception technology, to provide customers with leading planar and three-dimensional intelligent visual hardware and software products and solutions of artificial intelligence company, Including SDK, IP, chips, etc. The main products are the first generation of intelligent eye ViEye Ⅰ, the second generation of intelligent eye ViEye Ⅱ, Intelligent Stereoscopic Imaging SDK, ViEye is a 'Natural Learning' based on three-dimensional intelligent vision technology for robots And mobile devices developed collection, perception, cognition, choice, feedback, control and other features in one of the three-dimensional intelligent visual products.
Vizum former chief scientist at the World Exhibition company's third-largest mobile phone chip Communications, Dr. Burson-Marsteller Beijing Hualu audio-visual Institute Managing Director Xiao Jian Technology Co., Ltd. was founded, the company's core team are from well-known semiconductor company not only has a strong intelligent image processing theoretical basis, but also has a strong technical accumulation on stereo vision, object recognition, converged applications depth study: the team has a lot of successful development cases in the mobile phone baseband chip, networking, intelligent hardware and other areas involved in production of nearly 20 Chip, involving communications and multimedia, the cumulative total sales of hundreds of millions of pieces, creating a huge market value.
As a chip research and development of artificial intelligence + + intelligent systems provider business, Vizum adhere intelligent line, adhere to the theory and technological innovation as the soul of the company's development, independent research and development 'Natural Learning' intelligent stereo vision technology, and remain within the scope of the global leadership based on intelligent eye ViEye this technology developed specifically for mobile devices and service robots, which features the world's first, the performance goes beyond the international mainstream manufacturers in the three-dimensional binocular visual perception and cognition quantify, and has been successfully applied In three-dimensional face recognition, three-dimensional gesture recognition, robot vision and other products.
On May 22, 2017, Vizum completed a RMB50 million round A round of financing from the industry leader. This is the first time Vizum has completed the investment from Hengkun Ruijin in the end of 2016, mass production of the chip time to complete the second financing. Today, Vizum core team members are from well-known semiconductor companies, has successfully developed a large number of cases in the mobile phone baseband chip, networking, intelligent hardware and other fields, to participate in a total of nearly 20 Section, involving communications and multimedia, the cumulative total sales of several hundred million pieces.
In August, Beijing Weijing Intelligent Technology Co., Ltd. announced the signing of a cooperation agreement with Xianyang Hi-tech Zone: Establishing a robot intelligent eye application research center and production base in Xianyang High-tech Zone. On December 8, invested by Beijing Wei Jing Intelligent Technology Co., Ltd. Shaanxi Wei Jing Robot Technology Co., Ltd. grand opening in Xianyang.The project is expected to total investment scale will reach 200000000 yuan, Xianyang City and Xianyang High-tech Zone on Wei Jing robot in base construction and investment gave great support and investment.
Enjoy the trend of technology
Founded in 2014, Wyatt Trend Micro is a high-tech company that specializes in artificial intelligence applications of medical electronic sensors and physiology big data. It integrates the company's BioRF patented technology of non-contact arterial radar and integrates module / chip solutions for various physiological signal acquisitions. Combined with the cloud of biomedical models for intelligent analysis of physiological signals.The company launched a series of products including cuff-free blood pressure change tracking watch, abnormal heart warning wearable, systemic arteriosclerosis detector and many other personalized precision medical products .
Yue Yu Trend Technology founder and CEO Zhu Yudong has served for many years before the venture the United States Qualcomm semiconductor product management director, the United States a sharp news semiconductor global market director, the United States Bertone semiconductor senior technical market manager in China. Hong Kong University of Science and Technology holds a motor and electrical engineering A master's degree, a bachelor's degree in electronic engineering from Tsinghua University.
On August 28, Wyatt Trend Micro announced that it had completed the Pre-A round of financing of 10 million at the beginning of this round of financing, which will be led by VC Ventures, Beijing Challenge and Chenhai Capital. In 2015, Wyatt Trend also won the bid of Chenhai Capital Million angel angel wheel investment.
Yuan Ding audio (UnlimiterHear)
Yuan Ding International is a Taiwanese professional manufacturer of scientific aids, established in 1997 and has more than 100 patents on speech and hearing, which are specially designed to assist those with disabilities. Since 2008, Yuanding International has been researching hearing loss and how to improve listening comprehension, Measurement and frequency shift frequency technology, through different service platforms, to create MFA ecological environment to enhance listening ability .In 2015 YuanDing audio, ready to focus on consumer electronics audio market, the business model for the IP technology authorized suppliers , To provide listening algorithm program.
Yuan Ding audio technology core millisecond frequency shift algorithm (Millisecond Frequency Adjustment; MFA), mainly in 1.5 milliseconds to hearing impaired people can not hear the weak high-frequency subsonic frequency compression and frequency shift algorithm technology, there are still residual Hearing the middle and low frequency range, restore the true nature of the sound. Yuan Ding's listening algorithm is mainly used for two major industries, one Bluetooth assistant / auxiliary devices industry, the second is the Bluetooth headset industry. The remaining items (including China, the United States, Europe and other countries, has made 49 patents, the United States 8), in the licensing of the algorithm, there are several IC factory and Yuan Ding negotiate IP authorizations.
The Hearable overall market, which includes hearing aids, hearing aids, earphones & Earbuds, will be the fastest-growing wearable market, setting off more than $ 16 billion in 2020. Yuan Ding will hear 'Ear Clarity' , EarEQ 'Perfect Sound', EarSafe 'Secure Sound', Ear Command 'Intelligent Listening' applies to CSR and the Microchip Bluetooth platform for future AI voice control on mobile devices.
Yuan Ding audio by a number of retired science and technology co-founded by science and technology education in the United States had served as chairman of the board of directors Yang Guo Ping returned to Taiwan in 1997 to start a business, and the establishment of Yuan Ding International Development and Technology-aided cultural and educational foundation, venture dozen In recent years, he has been focusing on the design of electronic aids to help disabled people communicate with others through science and technology aids, during which hundreds of patents on hearing and acoustics have been accumulated. Vice Chairman Yang Xiaoming has served as the general manager of IBM Greater China PC Company, After serving NTT Technology and managing director of Winbond Electronics Shanghai, with more than 30 years of experience in the semiconductor industry, General Manager Huang Jinxing is in charge of business promotion. Will retire after retirement? Many self-managed retirees start to invest Another risk in the second half of life, Yuan Ding audio team that is a great case.
Baidu
Baidu United Spreadtrum and other hardware manufacturers launched the DuerOS smart chip can be seen as a new exploration of the integration of artificial intelligence and hardware devices in Baidu. DuerOS smart chip with low-cost chips and modules, the situation can be embedded in the chip into any hardware , Can be more quickly and widely applied to more scenes.It can be seen, Baidu is using the combination of 'algorithm + chip' industrialization of artificial intelligence landing.In March of this year, Baidu announced with the Zhaoyuan Zhaorui, ARM, Shanghai Hanfeng reached a strategic cooperation to build a three-tier structure including Miracle brain, voice solution and chip / module. The first two layers are provided by Baidu Miracle, which gives the chip DuerOS 'core function chip The modules are respectively developed by Ziguang Zhanrui, ARM and Hanfeng, the main body of the chip is Zhaopin RDA5981, which is a low-power WiFi + MCU for IoT applications based on ARM architecture and support WiFi, and supports hardware encryption acceleration.This chip does not support the acceleration of artificial intelligence operations, but can help to connect smart hardware (such as smart speakers, smart) to the network, And in the cloud to complete the artificial intelligence computing and interaction.
In addition, Baidu recently released XPU at the Hot Chips Conference in California, a 256-core, FPGA-based cloud computing acceleration chip that partners with Celestine. The goal of XPU is to balance performance and efficiency, And deal with a variety of computing tasks.
In addition to Baidu, Tencent and Ali also joined the war situation, have recently launched FPGA cloud solutions, Huawei is also doing cloud computing architecture and methods.
Clouds know the sound
The cloud is known as the most professional voice interactive provider, the company was founded in 2012. Headquartered in Beijing, in Shanghai, Shenzhen has offices and offices. At present, cloud-acoustic has launched a series of voice recognition service middleware and voice Recognition service cloud platform, widely used in mobile Internet, smart home, wearable devices, car navigation, medical, education, call center and other fields, for enterprises and users to provide professional voice recognition services.
Cloud aware highly integrated AI customizable chip UniOne, built-in DNN processing unit, compatible with multi-microphone, multi-operating system.In August 17, cloud knows sound officially announced a strategic investment of 300000000 yuan, the follow-up will be mainly in three The Company continued to exert its efforts in the following aspects: First, to strengthen R & D investment in key technologies of the IoT artificial intelligence service industry chain; Second, to step up R & D efforts in UniOne, a dedicated ASIC for artificial intelligence, to further improve product development and Commercial landing; Third, to further increase in the smart life, smart services industry chain ecological construction investment.
The cloud voice of the planning is from the early provision of open voice cloud platform services to follow-up focused Internet of Things, trying to transfer the technical capabilities from the cloud to the terminal, through AI core, AIUI side, AI Service three solutions to build 'cloud core' product strategy , Focusing on the layout of home, car, medical areas, thus forming a complete ecological closure.
Deep thinking
Think Deeply of AI, founded in August 2015, is a technology company that specializes in core algorithms for artificial intelligence (machine learning, NLP, NLM, machine vision) and hardware chip design. The main products are ideepwise AI Robot Service underlying service system, based on the underlying service system IDeepwise artificial intelligence 'medical brain' and second-generation artificial intelligence robot system and FPGA AI acceleration chip.In November 2015, Shen Pondering Xinjing Park has invested 1.6 million angel investment, and then In February 2016, it obtained 10 million preA round of financing of Joyoung.
The company's core team consists of senior scientists from the Chinese Academy of Sciences Institute of Automation, Software, Computing, Semiconductors and other institutions of artificial intelligence, machine learning direction, currently has 28 national invention patents, utility model patents 1, the trademark 1 13 software copyrights, the company and the Chinese Academy of Sciences Institute of Software, Chinese Academy of Sciences Institute of Computing, Chinese Academy of Sciences Institute of Automation jointly established 'deep learning artificial intelligence Institute' to promote the integration of production, research institutes.
As a driver of deep learning hardware technology, deep thinking of deep learning algorithms at the algorithm-level and chip-level depth optimization, providing unique deep learning chip solutions.Now available: FPGA + ARM Deep Learning Integration Platform Deep thinking Built ' FPGA + ARM deep learning integration platform ', which takes full advantage of FPGA's high-bandwidth reconfiguration and ARM processor application advantages.On this platform, CNN, DNN and RNN algorithms can be efficiently accelerated to improve the response speed of artificial intelligence devices And greatly improve the use of experience Deep learning calculation ASIC program Deep thinking for enterprise customers Deep learning computing ASIC (semi-custom integrated circuit) design integration program, in 65nm / 45nm / 40nm / 28nm and other advanced technology nodes to provide customers with unique Deep learning chip design support, the artificial intelligence experience to the peak of computing performance Deep learning algorithm + high-performance chip
Deep thinking Deep learning chip R & D team is breaking the limitations of traditional intelligence and constraints in this artificial intelligence competition to achieve wisdom in the algorithm and chip-level depth of integration.
In December, deep thinking released the third generation of artificial intelligence interactive robot iDeepWise.
West Well Technology (westwell)
Xijing Technology was founded in January 2016, focusing on the platform research and commercial application development of the brain-based intelligent chip + algorithm based on neuromorphic engineering. It has developed a self-developed deep learning brain chip (deepwell ) And deepsouth-like neuron-like chip products that simulate 50 million neurons.
West Well Technology has built a software and hardware platform for SNN pulse neural networks and deep learning - Westwe Brai. The first deep learning processor Deewe, it is learned that the power consumption is only a few one to several hundredth of the traditional chip .
Tan Li Min, CEO of Xijing Technologies, has said Westwe Brai has surpassed Google's Human Brain Simulator from the perspective of simulated neurons. The Google Human Brain Simulator uses a CPU connection while Westwe Brai gets out of von Neumann Computer architecture, using a circuit that simulates neurons.
In just a few moments, the startup has set up a team of postdocs, PhDs and masters graduated from the Polytechnic Associate in Artificial Intelligence, Imperial College London, the University of Oxford, Marseille and the University of California, Berkeley.
In June, Xijing Technology won Series A financing and investment from Fosun Tonghao. This round of financing will be used for technology research and development. Xijing Technologies obtained financing from Angel Wheel and pre-A round of investment in June 2015 and January 2016, respectively Parties include Heli investment, source investment, PreAngel and others.
According to the statistics of Tencent Research Institute, there are 14 enterprises in China's AI industry base (mainly processors / chips), and the AI era promotes the research and development of chips and various giants each consolidate their position in their respective fields of expertise , CPUs, GPUs and field programmable gate array chips are in full bloom. Under the huge engine of artificial intelligence, the semiconductor industry ushers in new changes. As the most basic and important part of the semiconductor industry, it will promote the rapid development of artificial intelligence , Complement each other community of destiny.
2. Why Taiwan's artificial intelligence may lose the mainland ?;
By 2020, every 3 mobile phones will have a built-in AI chip, but the AI chips that are now emerging are mostly new to the mainland, and why is Taiwan absent this time?
"I heard about the CPU, the GPU, and the NPU." At the end of November, Harmonic, the host, Ken Kennedy spoke at a press conference in Taiwan at Mate 10, the latest flagship handset for Huawei, with pre-arranged lines.
The mobile phone conference held at Dazhi MeiFu Hotel is full of subtle sense of incompatibility.A Ken and fashion singer Guo Shuyao and Xiu JieKai platform, the scene is a tough technical term.
▲ Huawei's flagship machine Mate 10 presentation, from the left repair Jie Kai, Guo Shuyao, host Ken Ken.
For example, NPU (Neural Processor Unit), Chinese can be turned into "neural network processor" is used to deal specifically with the depth of the learning function of the accelerator.
Or, the popular AI chip, the Kirin 970, the world's first AI phone, with a new AI kernel, so the user needs to manually set the shooting scene mode in the past and can now automate it - for example, When you focus on a steak, a knife and fork symbol appears on the screen, indicating that the AI processor recognizes it as food and automatically adjusts to the food's shooting mode.
Apple, Huawei leading AI attack terminal device
The world's third largest mobile phone brand, is showing the current popular semiconductor industry topic "AI on edge" (AI on the terminal device).
Apple's announcement in September this year also revealed that the A11 Bionic chipset on the iPhone X also features an AI-like "nerve-like engine" and, in one fell swoop, "AI on edge."
According to the past as long as the apple is used, it will become the industry "standard" practice, the next high-end handsets of various brands will also be equipped with an AI chip.
A month after Apple's announcement, Counterpoint, a technology market regulator, said in a research report that the market share of smartphones that target AI chips in 2020 will have skyrocketed from 3% in 2017 to 35%. Said that there is one AI chip for every 3 mobile phones.
"The AI is the terminal to come, Huawei and Apple are the first two," also attended the presentation Huawei TSMC IWTO Wang Yaodong, director of senior. And these two chips, using TSMC's most advanced 10 nanometer process.
The concept of the past, will think that a large number of mechanical learning needs computing, need to be done in the cloud, calculate and then return to the phone.
Wang Yaodong said that now the semiconductor process miniaturized to 10 nm, a substantial increase in computing power, enough to be the original cloud computing model shrink to the phone, but for the average general-purpose processor is still overloaded, so add Dedicated NPU. Its computing structure for a simple but a large number of parallel computing, "Wang Yaodong said.
A similar thing happened about 20 years ago when the demand for graphics operations skyrocketed, creating an independent new chip category, the graphics chip.
Huang Renxun, the founder of graphics chip maker Nvidia, first proposed the word GPU in 1999 to simulate the CPU (central processing unit) familiar to the public and let everyone know the importance of the graphics chip.
▲ Huida founder Huaren Founder called the father of GPU, and now he is also the latest AI chip battle grab the initiative, is expected to be listed later next year, the latest AI computing platform Pegasus, claiming to be the world's first to support the fifth Autopilot platform. (Source: Technology News)
Usher in contemporary "Cambrian explosion"
Now in response to AI's demand for the explosion of computer performance, a new chip category has already emerged, but what is it? At present, the industry is incomprehensible, Google called TPU its own server, Huawei called NPU, Intel called NNP (Neural NetworkProcessor).
"This is a big trend," said Zhou Li Dong, vice president of Microsoft Research Asia, which specializes in artificial intelligence systems. "Now it's the XPU era. There are many blooms and lots of solutions come up."
Many people in the industry use the famous Cambrian explosion known as the history of paleontology to describe the lively billowing look now.
Interestingly, the current China's top AI chip design company, called Cambrian technology.The company funds and people team are from the CAS Institute of Computing, by the Chinese media as "AI team" and in August this year Rapid completion of the 100 million US dollars A round of raising funds, shareholders, including Alibaba, Lenovo, although the valuation has not been exposed, but the industry generally believe that the Cambrian has become the world's first AI Unicorn Unicorn field (valued at more than 10 One hundred million U.S. dollars) .
Because, Huawei AI processor Mate 10 core, is from the Cambrian authorization.
"The first option is quicker in the market," said Wang Yaodong. However, he believes Huawei will definitely find ways to use its own AI design next. "Anyone who has the ability to develop his own AI will develop on his own."
▲ Cambrian science and technology originated from the Chinese Academy of Sciences in China, which is Miao Honggen's "AI hope." Perhaps the founding team received very little media coverage.
New breed of processor
The end of October, the global semiconductor industry leaders gathered 30th anniversary of TSMC forum, AI chip has also become a hot topic.At that time, UBS analyst Lu Jia 璈 from the audience asked: "The future prospects of AI chip? Like GPU, DSP become the mainstream IC Category? "
At that time, TSMC chairman Zhang Zhongmou appointed by the most suitable for the question Huang Renxun answer.
Jen-Hsun Huang believes there will be 3 different types of AI chips, the first of which is used to train deep learning models that are computationally powerful and often placed in cloud servers.
The second is the neuro-processor used on Huawei's mobile phones and Apple's mobile phones. "According to Jen-Hsun Huang," a new breed of processors, all future autonomous machines, including unmanned vehicles, unmanned aerial vehicles, People) will have one on top.
Third, the endpoints used in the Internet of Things, close to the sensor, the number of megabytes calculated above.
"In the future there will be AI and mechanics learning on every sensor," added Vincent Roche, chief executive of analog IC maker ADI. "Because a lot of computation needs to be done close to the source of the data."
So many new semiconductor needs, it is also the price of TSMC all the way this year exceeded 200 yuan, the market value of Taiwan's currency soared 6 trillion yuan key.
No wonder, foreign brokerages have recently lowered their TSMC target prices.Zhang Zhongmou can calmly tell the media after attending the breakfast meeting of the AIC. "At present, everyone is talking about AI. AI needs high-speed calculation and development fundamentals Can not see any restrictions ....... (efficient computing) There is no need to slow down the worries. "
Taiwan's newly emerging menace is absent instead?
Currently the most watched in the market, Huida, Intel and many Cambodian-led start-up companies in China, are competing for the UAV, UAV, smart phones and cloud server high Performance AI chip.
Taiwanese manufacturers are still absent from this fertile virgin land.
A senior industry revealed that such high-end chips, the need for R & D team, the use of the process costs are not out .Also go out with Intel, Huida fight .There have been a well-known manufacturer of the team tried to develop, but still in this year Played back, because I think the risk is too great. And even venture capital does not support.
In contrast, China's newly created team, although less skilled and experienced than Taiwan, said, "China now has so much money to raise it," said Lu Chaoqun.
▲ semiconductor veteran, Yuchao Technology chairman Luchaoqun said that in the face of the mainland government's strong investment in new technology of AI, Taiwan's IC design industry still needs the support of all sectors of government.
What's more, the mainland government announced in July that "a new generation of artificial intelligence development plan" will call for achieving the goal of synchronizing AI technologies and applications with the advanced world standards by 2030. It is expected that there will be a new wave of hot money To the AI industry.
"The funds invested in the Mainland are far worse than manpower and material resources. Taiwan has to chase it," Wang Yaodong could not help but say.
3. 'Lead' artificial intelligence: AI usher in a new era in China;
Artificial Intelligence (AI) should be selected as the hottest topic in science in 2017. The supportive layout at the national level and the comprehensive breakthroughs in the industry mean that artificial intelligence may still be capital-competitive in 2018 and even in the next decade or two decades tuyere.
On December 14, 2017, the Ministry of Industry and Information Technology released the "Three-year Action Plan to Promote the Development of a New Generation of Artificial Intelligence Industry (2018-2020)" (hereinafter referred to as the "Three-year Action Plan"), followed by Zhejiang Province in 2017 December 26 issued a new generation of artificial intelligence development plan.Beijing on December 26, 2017 also issued a document to speed up scientific and technological innovation to build high-tech economy structure series of documents, involving artificial intelligence and other 10 industries, and separately prepared guidance .
The industry is bustling and has made substantial breakthroughs in AI chips, operating systems, applications, etc. The Cambrian launched three AI IP processors on November 6, 2017, with Horizon in 2017 December 20 also launched two artificial intelligence embedded processor products.Is committed to operating system to reduce the threshold of AI Baidu, the fourth paradigm, through the smart speakers, AI phones and other specific products affect the user Alibaba, Jingdong, Huawei, Lenovo, etc., in 2017 has also made some progress.
Chinese Academy of Engineering Pan Yunhe told the "China Business" reporter that artificial intelligence is the core technology of the next generation of science and technology revolution. It is bound to promote a new generation of industrial revolutions. China must make full preparations, seize the opportunity to achieve lane development, and overtake corners.
Start synchronously
PC Internet era, almost all the world with Intel and AMD chips, Microsoft Windows is at the operating system level 'dominate the world.' Mobile Internet era, in the field of chips that Qualcomm means high-end market, China MediaTek, Spreadtrum You have been hoping to catch up, but only in the low-end market has a certain role; the operating system level is still iOS and Android world.In the artificial intelligence era, it is another story, where the chip, operating system, application level, Chinese companies and global players are all on the same starting line.
For example, in the field of chips, although the GPU of NVIDIA has taken the lead, there are not many AI chip schools such as Intel, CPU + FPGA, Intel, TPU, TPU, AMD, Qualcomm, etc. But far from deciding the outcome, Including the Cambrian, Horizon, Vimicro, Xijing Science and Technology and other Chinese 'players' are also expected to win in the field of AI chip belongs to China's place.
Among them, Xijing Technologies launched neuron-like brain chip in 2016; Vimicro also introduced the first embedded neural network processor (NPU) in China in June 2016; and Cambrian released on November 6, 2017 for terminal and cloud Of the new generation of artificial intelligence processor; Horizon on December 20, 2017 also introduced the first generation of self-developed Gaussian architecture artificial intelligence vision chip and announced from 2017 to 2019 year by year architecture - Gaussian architecture, Bernoulli architecture, the speed of Bayesian architecture, constantly iterating on their own artificial intelligence chips.
Although the volume on the market is far less than that of international players like Google and Nvidia, Chinese players are not lagging behind in the field of AI chips. "Horizon was the first in the industry to propose to create an artificial intelligence chip architecture BPU, Google put forward the TPU architecture earlier than the horizon. 'Horizon founder and CEO, the state' Thousands of people 'plan Distinguished experts, the Ministry of Science and Technology National Advisory Committee of a new generation of artificial intelligence, Dr. Kai Kai told this reporter.
Yu Kai served as vice president of Baidu Research Institute, director of Baidu Depth Learning Lab, is one of the AI technology resign from Baidu.Kai Kai revealed that in 2015 he decided to leave from Baidu, that is, to create AI 's China Core. "Yu Kai believes that the AI chip at the national level relates to whether China can overtake a corner in the era of artificial intelligence, and the proprietary AI chip in the industry also means that algorithms such as deep learning are not limited by the computing power in the future.
Chen Tianshi, founder of the Cambrian analysis, 'GPU is the current mainstream, but its basic structure is not designed for AI, the efficiency is subject to many limitations.Although the FPGA iteration fast, short-term to meet certain computing needs, but the speed and energy consumption ratio , There is a gap with the proprietary AI chip.The ideal AI chip should be a new processor that can adapt to a wide range of applications such as voice, image, video, natural language multimodal processing, while far beyond the CPU and GPU Efficiency. "Chen Tianshi also believes that China's 'players' are expected to use AI chips to achieve corner overtaking.
The same is true of the operating system field.Related estimates show that although the Internet industry in China is also well-established in the world, at the operating system level, it takes approximately several hundred billion yuan each year to deliver to Windows, iOS and Android operating system giants Interests. The advent of the artificial intelligence era, also means that the operating system usher in the reshuffle opportunities.A few days ago, Baidu developed the interactive AI operating system Duer OS has been successfully selected the Fourth World Internet Conference 14 leading scientific and technological achievements list, Is regarded as the domestic market with very promising AI operating system.In addition to Baidu's Duer OS, the fourth paradigm developed by the 'Prophet' system, but also made it clear that to become the Windows of the AI field.
'In the PC Internet era, Windows drastically reduced the difficulty of using PCs for ordinary people. In the AI era, the Prophet system also wanted to do something similar and lower the threshold for using AI. "Dai Wenyuan, founder of the fourth paradigm, told Ben Reporter.
Country driven
In addition to the basic fields such as chips and operating systems, innovation in AI applications is also emerging in China, and besides the overall layout of BAT in the field of artificial intelligence, there are a large number of future stars in the application of artificial intelligence in China, such as the voice field IFT fly, go out to ask, think of Chi and so on, the vision of the Shang Tang technology, Kuangsheng science and technology.
Liu Feng Feng, founder of iFLYTEK, said openly that the application of artificial intelligence in China has been on the same starting line with the United States. With the help of state policies, the advantage of applying artificial intelligence in China is actually more obvious.
Lei Jun, the founder of millet, had previously said in an interview with this reporter that China has the advantage of developing artificial intelligence in a huge talent pool and huge amounts of big data. Relevant statistics show that from 2006 to 2016, the Chinese contributed about 30% of the top global workers Lei Jun believes that the Chinese people attach importance to mathematics since childhood, good science and engineering base, which in the age of artificial intelligence is a huge advantage.At the same time, China has 730 million Internet users, living in the world, has accumulated a large number of Internet users behavior data. Di think tank predicts that by 2030, China's total Internet data will replace the United States as the world's first, which will also become China's development of artificial intelligence unparalleled advantage.
Chinese Academy of Engineering Pan Yunhe told this reporter, looking at the developed countries in the world of artificial intelligence concerns and attention, the United States in October 2016 issued a "National Artificial Intelligence R & D Strategic Planning", the United Kingdom in December 2016 released "Artificial Intelligence: The Future Opportunities and Implications of Decision Making. "In comparison, the State Department of the People's Republic of China listed" Internet + Artificial Intelligence "as one of the" Guiding Opinions on Actively Promoting the Internet + "released in July 2015 One of the key actions.
Subsequently, the State Council issued "A New Generation of Artificial Intelligence Development Plan" on July 20, 2017, clearly setting the goal of "three steps" for China's artificial intelligence. The first step is to establish a preliminary technical standard, service system and industrial ecology by 2020 Chain, cultivate a number of the world's leading backbone enterprises, the core industry scale of more than 150 billion yuan, driven by the industrial scale of more than 1 trillion yuan; the second step is to 2025 in smart manufacturing, smart healthcare, smart city, smart agriculture, national defense construction And other areas of extensive application of artificial intelligence, the core industry scale of more than 400 billion yuan, driven by 5 trillion yuan; the third step is to 2030 in the production and living, social governance, national defense construction and other areas to greatly expand the breadth and depth of the formation of core technologies covering , Key systems, support platforms and smart industrial applications. The core industries are over 1 trillion yuan in scale, bringing in 10 trillion yuan.
In the "Three-Year Plan of Action," the MIIT explicitly put forward the proposal of "speeding up the development of the artificial intelligence industry and promoting the integration of artificial intelligence and real economy" and explicitly put forward the design, foundry and packaging technologies in the areas of smart sensors and neural network chips International advanced level and other four key development goals, as well as in the smart car network, intelligent service robots, smart unmanned aerial vehicles and other eight key areas of large-scale applications, the formation of a strong global competitiveness.
It is generally accepted by the insiders that domestic enterprises that have already emerged in the field of artificial intelligence will be the first to benefit from the guidance of national and provincial planning that explicitly aims to cultivate the world's leading or the nation's leading artificial intelligence enterprises.
4. Year of artificial intelligence eruption: lack of capital influx of professionals cited foam concerns
BEIJING, Dec. 30 (Reporter Qiu Yu) 2017 is a big year for artificial intelligence (AI), a concept that is rapidly warming up in the world and attracting a lot of talents in China Capital inflows, all kinds of enterprises and applications blossom everywhere, but at the same time, the chase of capital also caused some bubble concerns.
Artificial intelligence usher in the year of the outbreak
Artificial Intelligence was foreseen by the 2016 World Economic Report as the core technology representative for the Fourth Industrial Revolution, which Bill Gates calls the 'Holy Grail' of the computer science community.
In March 2016 and May 2017, the two human-computer war games in the Go community brought artificial intelligence to more people's perspectives. AI programs Alpha Dogs successively beat the top players, Li Shishi and Ke Jie, This dog 's first top butt has not sit hot, in October 2017, a new generation of Alfafar out of the mountain, without any chess, with 100 to 0 over the elvis dog.
However, artificial intelligence is much more than Go-Go, and it has become more widely used in areas such as speech recognition, face recognition, driverlessness, big data analysis and even disease prediction. All aspects of life.
China's AI field is also sparkling. Baidu has been betting artificial intelligence, driverless cars on Beijing's Fifth Ring Road, and announced that driverless technology is free; Jingdong, Ali, Xiaomi also actively to the artificial intelligence layout, involving Water smart speaker market; Internet giant Tencent released an artificial intelligence development strategy.In addition to the giant, Horizon, Cambria and other start-up AI companies are also very popular.
In fact, the concept of artificial intelligence as early as the 50s of last century has already been born. Why the rapid development of the trend in recent years?
For example, Lu Fan, founder of the National Thousand Talents Program who has been working in this field for many years, said that in recent years, the advent of deep learning has driven the emergence of artificial intelligence. "Deep learning has made algorithms more efficient and machines We can abandon the laws that people grasp and dig deeper into the essence of things. '
Deep learning is a machine learning method that rapidly evolves as the computing power of the computer increases dramatically, akin to the work of the human brain nervous system.
In addition, after decades of development, the chip has been able to support the AI algorithm with very high computational complexity. "Lu Fan said that big data is also a key factor in the development of artificial intelligence, which can give feedback to the machine, keep on training, and tell the machine that it is right And wrong '.
Talent demand explosive growth
To make a program as smart as a human being, first and foremost the need for so many smart people, as the concept of artificial intelligence heats up, so does the need for talent.
Researcher at a state-level chip design institute told reporters that this year is generally felt that newcomers are more willing to do artificial intelligence and big data.He disclosed that a domestic science and technology giant company's infrastructure department only received two resumes on campus this year, Hundreds of resumes in the startup's AI division stand in stark contrast.
Zhou Yang, who has been working in the traditional chip industry for nearly 20 years, chose to switch to an AI startup that is engaged in the design of embedded AI chips, and in his opinion, AI chips are now designed to focus on smart drivers, smart cities Application scenarios such as landing, there is more room for development.
Despite the large number of applicants, there is still a large talent gap in the field of artificial intelligence.
According to a report released by Zhaopin, the demand for artificial intelligence personnel nearly tripled in the past year, and the growth rate of algorithm engineers was the most rapidly, and the personnel with AI skills now mainly focus on the monthly salary ranges of RMB1,000 to RMB15,000, accounting for 40%.
Luo Huan, the founder of Dana Technology, has 20 years of experience in software engineering and artificial intelligence, and said that excellent algorithm engineers are of great help to the development of the company and the company is willing to hire a capable R & D personnel for hundreds of thousands or even millions of annual salary. Due to the lack of qualified personnel in this area, professors and researchers in academia are being besieged by the industry.
Globally, according to the "Global White Paper on AI Talents," jointly released by Tencent Research Institute and BOSS Direct, the total number of global AI professionals is about 300,000, while the current market demand is in the order of one million, while AI talent supply exists Big gap.
Capital influx cited foam concerns
Talent and funding are key factors in the development of the industry, and artificial intelligence is emerging as a new territory for capital chasing.
In July 2017, Shang Tang Technology was awarded $ 410 million of round B financing, setting a single-round financing record for the then global artificial intelligence field. In August, Cambrian Technology Inc., An artificial intelligence chip company hatched from the CAS Institute of Computing, completed 100 million U.S. dollars A Round financing, becoming the world's first unicorn AI chip field start-ups.
In October, Kuang as Technology announced the completion of 460 million USdollars C round of financing, refresh the previous Shang-Tang technology to maintain the record; the same month, Horizon received about 100 million US dollars of Intel A + round of financing; December, Tu Lin Technology won 250 million yuan A round Financing.
The influx of funds, more and more companies began to affix artificial intelligence tags, triggering concerns about the bubble of artificial intelligence.
In March of this year, Kai-Fu Lee openly told the media that there was a lot of bubbles in investment and project valuation in the field of artificial intelligence, and many pseudo-artificial intelligence products appeared. 'I met an underwear maker and also said that I am an artificial intelligence enterprise It is a very abnormal phenomenon.
'Next year will certainly be eliminated a number of artificial intelligence companies.' Luo Huan said that now, really understand artificial intelligence, very few investors, few investors can correctly evaluate the technical content of a project, but investors learn fast, Artificial intelligence companies without core strengths have less access to finance.
Lu Fan believes that excessive capital inflows will inevitably lead to the formation of a bubble, but the bubble does not mean that the industry is false. Just as there was a bubble in the rise of the Internet, the general direction of the Internet was right and the wave went , The remaining have become the backbone of the force.
Can China overtake a corner?
Compared with western countries, artificial intelligence seems to be more sought after by Chinese capitalists, some analysts said that China's big data, demographic dividend, policy support and other advantages is an important reason.
From a policy perspective, in March this year, 'artificial intelligence' was first written into the government work report. In July, the State Council issued the "New Generation of Artificial Intelligence Development Plan." By 2030, the scale of artificial intelligence core industry will exceed 1 trillion In November, the Ministry of Science and Technology launched major projects and announced the four major platforms of artificial intelligence. In December, the Ministry of Industry and Information Technology released a three-year plan of action to strive for a breakthrough in artificial intelligence by 2020 and form an international competitive advantage in several fields.
According to a report released by Eurasia Group, data size is the most important factor in the competitive edge in the field of artificial intelligence. China possesses vast amounts of data far beyond other countries.
The report said that while overall China's artificial intelligence lags behind the United States as a whole, it will shift from imitation to catch-up and lead or jointly lead the world in terms of data advantages, capital support, government policy support, talent expansion and hardware manufacturing advantages. Wave of artificial intelligence.