The age of intelligence is coming sooner or later. Each era has its core material carrier, such as the steam engine in the industrial age and the general-purpose CPU in the information age. This core carrier will also appear in the intelligent era. What the company wants to achieve in the future is to increase the computational efficiency of artificial intelligence chips by 10,000 times and reduce power consumption by 10,000 times. In the era of geology known as the "Cambrian" about 600 million years ago, a large number of invertebrates appeared "life explosion" in a short time. Nowadays, the name "Cambrian" has once again been mentioned. It is derived from the naming of artificial intelligence chip processors developed by the Institute of Computing Technology of the Chinese Academy of Sciences, which means that artificial intelligence is about to usher in an era of great explosion. "Cambrian" won a $100 million financing from Alibaba in August, becoming a unicorn company in the smart chip field with a valuation of more than $1 billion. The founders Chen Yunqi and Chen Tianshi’s brothers also made a “blockbuster†and jumped into the public eye. At present, the Cambrian terminal processor IP products have been derived from 1A, 1H and other models. In the next few years, hundreds of millions of terminal devices around the world are expected to achieve powerful local intelligent processing capabilities by integrating the Cambrian processor. . Chen Yunxuan's goal is to increase the computational efficiency of artificial intelligence chips by 10,000 times and reduce power consumption by 10,000 times. In order to achieve this goal, in addition to the rapid development of capital with the help of capital, it is also necessary for Chen Yunqi and Chen Tianshi to continue their long-standing "double swords." At the Innovative Conference called "Explorer" held by the Chinese Academy of Sciences at the end of August, his brother Chen Yunxi appeared, and he was accompanied by many highly respected old academicians, including the Institute of Biophysics of the Chinese Academy of Sciences and the Shanghai Institute of Bioscience. Academician Guo Aike, a researcher at the Institute of Science, and the chief scientist of Quantum Satellite, and Professor Pan Jianwei of the Chinese University of Science and Technology. Chen Yunxuan wore a white round neck T-shirt with the Cambrian word, and it was a casual dress, just like the entrepreneurs in Silicon Valley. He has a work card and a key around his waist, and he holds a computer in his hand. At first glance, he is an uncompromising science man. Sun Liguang, a professor at the School of Earth and Space Science at the University of Science and Technology of China, who is sitting next to the First Financial Journal, explained: "The cloud is a typical Keda. He is not paying attention to this dress. The generals are so worn, so don't blame him. ." Chen Yunxuan used the title "Make Machine Brain" as the title of his speech. The "genius boy" who was selected as one of the world's 35 best innovators under the age of 35 under the MIT Technology Review in 2015, said: "What the company wants to achieve in the future is to increase the efficiency of artificial intelligence chips by 10,000 times. The power consumption is reduced by 10,000 times. This means that we can put things like AlphaGo on the phone, let the phone help us do all sorts of things, and even realize strong intelligence through long-term observation and learning." The reason why he called it a "genius boy" is because Chen Yunqi started to go to middle school at the age of 9 and entered the junior college class at the age of 14 at the age of 14. The younger brother Chen Tianshi also basically followed his brother's growth path. "I have been doing chips since I graduated. My brother has been doing algorithms, and the chip plus algorithm has just produced the 'Cambrian' artificial intelligence chip." Chen Yunxi told reporters. Chen Yunqi is a native of Nanchang, Jiangxi. His father is a power engineer and his mother is a history teacher. This also cultivated the characteristics of the two brothers "both liberal arts and sciences". Chen Yunqi is obsessed with history and especially loves reading, whether it is engineering or historical. Chen Yunqi, who is now the father of twins, believes that interest is very important for children. Chen Yunxi loves to play games. He is quite proud to tell reporters that his undergraduate major is mainly engaged in "StarCraft." But also in the game, Chen Yunqi got the inspiration for the chip. He said that he is very happy to see AI giant training machines such as Facebook and Google's DeepMind, challenging the human players of StarCraft. AI researchers can now build their own models using open tools to address the technical challenges of StarCraft. Before the founding of the Cambrian, Chen Yunyu participated in the development of China's first general-purpose CPU chip, the Longxin No. 1, in the last year of the university. This was a glorious and rare opportunity for him. In 2002, he came to the Institute of Computing Technology of the Chinese Academy of Sciences, and followed the research of Hu Weiwu, a researcher, and became the youngest member of the Loongson R&D team. After graduating from the Ph.D., he stayed at the Institute. At the age of 25, Chen Yunyi has become the main architect of the 8-core Godson 3. The Institute of Computing Technology of the Chinese Academy of Sciences has trained a large number of high-tech talents for the Chinese computer industry and academia, and has gone out of a number of high-tech enterprises such as Lenovo and Shuguang. It is also an important shareholder of Cambrian Technology and a long-term partner of industry, education and research. Liu Chuanzhi, the founder of Lenovo, was a researcher at the Institute of Computing Technology of the Chinese Academy of Sciences. Later he turned one of the labs into the predecessor of Lenovo. After Lenovo’s profit, the feedback from the calculation office was very large, and it also helped build the new building. From "Godson" to "Cambrian", Chen Yunqi and Chen Tianshi aimed at the ecological layout of artificial intelligence. He once explained his source of inspiration: "The human brain is the most intelligent object in the known world. If you can digitally abstract neurons and synapses in your brain, such a digital network may inherited people to some extent. The ability of the brain to process information." However, although neural networks are a good method for intelligent processing, the high power consumption and low efficiency of general-purpose processors have hindered the development of artificial intelligence chips. Chen Yunqi explained: "If you want to build a human brain-scale synaptic neural network with a general-purpose processor, you may need to build a power station to power it. AlphaGo uses 1000 CPUs and 200 GPUs per minute. It is as high as $300, and the network is only one-thousandth of the human brain." The Cambrian AI chip solves this problem precisely – it simulates the calculation of neurons and synapses in a computer, intelligently processes information, and manages 16 billion per second by designing specialized storage structures and instruction sets. Neurons and more than 2 trillion synapses, power consumption is only one-tenth of the original, and even hope to put the entire AlphaGo system into the phone in the future. In Chen Yunqi's view, the creation of a machine brain with human intelligence means that human beings will be freed from tedious manual labor and simple mental work, focusing on creative activities. For Chen Yunqi, who has a pair of twins, it can be said that it can solve the immediate needs of the eyebrows. Qiu Zilong, from the Institute of Neuroscience of the Shanghai Institute of Biological Sciences, Chinese Academy of Sciences, said to Chen Yunqi: "If you have the machine brain you are looking for, what is the happiest and most unhappy thing for you?" Chen Yunxi responded humorously: " The happiest thing is that someone helped me with my child. The most unhappy thing is that I have no children." Academician Pan Jianwei asked Chen Yunqi: "Since the machine can learn quickly, can you directly copy the things learned by the machine into the human brain? If a machine has been trained to learn something, can it be copied to other machines?" In this regard, Chen Yunxi said that scientists now manipulate virtual neurons in neural network chips, rather than real neurons. To copy data into the human brain, it is still not realized. However, it is now possible to copy a machine that has already been learned to another machine, and the cost of such copying is relatively low, which is a more feasible and cost-effective method. In addition, in response to the recently proposed "AI-humansymbiote" by Tesla founder Musk, Chen Yunqi believes that this "digital expansion of the human brain" external equipment is still difficult to achieve. He told the First Financial reporter: "This is still very technically difficult. Now humans can only copy some of the relatively shallow information in the brain." But looking forward to the development of the brain-computer interface in the next decade, he said: "This is possible. Many people are now trying to do this." For the future development of artificial intelligence, Chen Yunqi and Chen Tianshi believe that as society gradually transitions from the information age to the intelligent era, artificial intelligence chips will be an indispensable carrier for intelligent computing. Complex deep learning networks require high computing power and require more and more powerful computing resources. GPU is the current mainstream artificial intelligence computing platform. Because its basic framework is not designed for artificial intelligence, efficiency is limited. Although the FPGA is iteratively fast, there is still a gap between the calculation speed and the energy consumption ratio compared with the dedicated artificial intelligence chip. In an interview recently, Chen Tianshi revealed: "There are still many companies and universities that are also tracking our previous achievements and developing ASICs for deep learning, such as Google TPU." He said: "Since the last century, the rapid development of information technology in the United States has emerged as a series of great chip companies represented by Intel, which has driven the transition of human society from the industrial age to the information age. In recent years, along with artificial intelligence technology and brain science. With the accelerated development, smart technologies represented by smart phones, smart driving, intelligent manufacturing and robots are beginning to mature. Human society is in the turning point from the information age to the intelligent age. The mission of the chip will be transformed from the calculation of the information age to the supporting machine. Intelligence. And Cambrian is the company that can take on this glorious mission." However, a professional from Nvidia told the First Financial reporter: "The Cambrian and NVIDIA chips are not very comparable because it is an integrated circuit (ASIC) designed for specific purposes, just like Google's TPU chip is Designed for Tensor-flow, there are limitations. NVIDIA's GPUs are more flexible, can be programmed according to different application needs, optimized for different algorithms, and can withstand the test of technology upgrades." In fact, ASICs are being used more and more because of the benefits of customization and low power consumption. Take the current hot bitcoin "mining" as an example. In the past, NVIDIA chips were used, but NVIDIA's GPU has a fatal weakness, that is, the power consumption is very large, and the "miners" need to pay high electricity bills. The cost will dilute their profits. To this end, bitcoin "miners" are increasingly turning to ASIC ASICs, which is a chance for companies like Cambrian, and on the other hand is a threat to NVIDIA GPUs. The rapid development of artificial intelligence is largely due to the improvement brought about by deep learning. Deep neural networks mean a huge amount of computation, and fast iterations require speed. As an artificial intelligence and a new neural network chip, in addition to the IBM brain brain chip TrueNorth, which currently integrates millions of neurons, the Cambrian artificial intelligence chip DianNao is also listed as one of the advanced chips. However, it is not easy for neural network chips to go out of the lab and enter the market. Yann LeCun, the father of the Convolutional Neural Network and Facebook's Artificial Intelligence Lab (FAIR), once commented negatively on IBM TrueNorth. In this regard, Chen Yunyi said in an interview two years ago that the neural network processor is in the "Spring and Autumn Warring States" period. This emerging field is different from the general-purpose CPU. There is not much difference in historical accumulation. "The opposite depth of neural network processor China is still the earliest, completely ahead of the possibility." Chen Yunxi said. At the end of last month, when the First Financial Reporter asked, in which industries the Cambrian chip will be used in the future, Chen Yunxi made it clear that there will be an answer in the next month or two. On September 2, Huawei released the world's first Kirin 970 mobile computing platform at the IFA show in Germany. Although not publicly announced, the AI ​​chip behind it is from the Cambrian. A Huawei senior executive told the First Financial Reporter that Huawei Haisi and Cambrian are indeed co-developing. Another media broke the news that the Kirin 970 integrated NPU (Neural Processing Unit, neural processing unit) concept has been brewing five years ago. In response to the cooperation with Huawei, Chen Yunqi did not answer the question of the first financial reporter, but his brother Chen Tianshi once revealed the commercialization of the Cambrian chip market: "The main two aspects are: one is the terminal, the other is the cloud. The terminal Products are mobile phones, smart glasses, bracelets, etc., which require chips to identify images, audio and video, and in the cloud, well-known cloud customers like Keda Xunfei and Shuguang are already Cambrian customers." Chen Yunyi said frankly: "Intelligence has evolved to the present, there are many advances in algorithms, and it can solve many practical problems, such as pattern recognition. But this is still far from the exciting intelligence that people expect. However, he believes that hardware research, especially neural network chips, will play a key role in the advancement of artificial intelligence, especially for the realization of advanced intelligence capabilities. The Cambrian has a very foreign English name called Cambricon. But in comparison, the name of the Cambrian chip is named after the very straight white Pinyin DianNao. It is the "computer" of Chinese Pinyin. It has the name of this grounded gas, and thanks to a Frenchman. Chen Yunqi said that the topic of neural network processor is a Sino-French cooperation project. Initially they remembered an English name similar to "Electricmachine" or "ElectricBrain". But at the time, a Frenchman from the team, Olivier Temam, a researcher at the French National Institute of Information and Automation (Inria), suggested to them that instead of taking a plain English name, it would be better to use the Chinese pinyin to name it. This is a "foreign language" for foreigners, but they will feel very "foreign". Surprisingly, the average age of members of the Cambrian team is only 25 years old, but most of them are “old drivers†of chip design development and artificial intelligence research. Many key members have begun to work in related fields during their school days. According to the forecast data of consulting firm TracTIca, by 2025, the market share of deep learning chips related to artificial intelligence will increase from US$500 million last year to US$12.2 billion, with a compound annual growth rate of over 40%. "The era of intelligence is coming sooner or later." Chen Yunqi said, "Every time has its core material carrier, such as the steam engine in the industrial age, the general-purpose CPU in the information age, and this core carrier will emerge in the intelligent era." The grade of optical round lens produced by our company can reach G5 grade, the sphere diameter ranges from 0.4mm to 250mm, and the sphere diameter tolerance can reach +/-0.001mm. The product consistency is good.
Fused Silica Ball Lens,Objective Ball Lens, Endoscope Lens,Glass Fiber Optic Ball Lens Bohr Optics Co.,Ltd , https://www.bohr-optics.com
Processing materials mainly include optical glass, quartz, float glass, high borosilicate glass, laser crystal, sapphire, ruby, germanium glass, germanium single crystal, zinc selenide, single crystal silicon, K9 crystal glass, calcium fluoride, silicic acid Salt glass, zinc selenide and cyanide infrared glass lens.
Among them, quartz glass is a special industrial technical glass with a single component of silica. This kind of glass has a hardness of up to seven on the Mohs scale, has high temperature resistance, low expansion coefficient, thermal shock resistance, good chemical stability and electrical insulation properties, and can transmit ultraviolet and infrared rays.
Small ball lenses are widely used to couple optical fibers and semiconductor lasers, or another optical fiber, or light detectors.