In the field of artificial intelligence, the basic abilities of language model, graphic model and even multimodal model of artificial intelligence have been fully demonstrated. For example, Ali Dharma Institute announced the latest progress of the multi-modal large model M6, and the parameters jumped from trillion to 10 trillion; Pengcheng Lab and Baidu jointly released the world's first 100 billion-level knowledge enhancement model-Pengcheng-Baidu Wenxin, with a parameter scale of 260 billion.
Not only that, but the cross-integration of artificial intelligence and other scientific fields has also sparked a spark. In the list of scientific breakthroughs in 20021year recently published by Science magazine, two technologies for predicting the structure of protein based on artificial intelligence, AlphaFold and RoseTTA-fold, topped the list.
In the field of human-computer interaction, when Zuckerberg renamed Facebook "Meta", Tesla and elon musk, CEO of Space Exploration Technology Company, focused on the brain-computer interface. Musk believes that brain-computer interface devices will be more likely to change the world and help people with quadriplegia or physical defects to live and work better. "Complex brain-computer interface devices can completely immerse you in virtual reality." In addition, in May this year, Stanford University developed an intracortical brain-computer interface system, which can decode the handwritten actions imagined by paralyzed patients from the neural activities in the motor cortex and convert them into texts.
In the field of supercomputing, it is most worth mentioning that in June this year, 5438+065438+ 10, China supercomputing application team won the "Gordon Bell Award", the highest international award in the field of high-performance computing applications, for its achievements in "real-time simulation of very large-scale quantum random circuits".
In terms of open source, RISC-V open source instruction set and its ecology have risen rapidly; The open source community of openEuler operating system, led by Huawei and participated by Institute of Software of Chinese Academy of Sciences and Kirin Software, has gathered 7,000 active developers, completed more than 8,000 self-maintenance open source software packages, and spawned more than 10 commercial distribution. ...
Looking back at 20021,the information technology edition invited industry experts to sort out the development context of the above four fields and look forward to the future development trend.
Author Zhang Shuanghu
AlphaFold is the "first brother" in the field of 202 1 artificial intelligence (AI).
Recently, Science magazine published the list of scientific breakthroughs in 20021year, and two technologies based on artificial intelligence, AlphaFold and RoseTTA-fold, ranked first.
A few days ago, AlphaGo and AlphaFold also made the list of "202 1 Top Ten Engineering Achievements in the World" selected by Journal of China Academy of Engineering.
In an interview with China Science, several experts talked about AlphaFold when reviewing the achievements in the field of artificial intelligence this year.
"The scientific discovery of AlphaFold and the artificial intelligence development ecology that China is building cannot be said." Wu Fei, director of the Institute of Artificial Intelligence of Zhejiang University, told China Science Journal.
Wang Jinqiao, a researcher at the State Key Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, nominated "Diagnosing COVID-19 with AI", "AI for Science" and "Zidong Taichu".
In the medical field, artificial intelligence has long been used to detect diseases such as pneumonia, asthma and Alzheimer's disease. MIT researchers have developed an artificial intelligence model that can identify COVID-19 patients by analyzing cough records. The accuracy rate of identifying cough in COVID-19 patients is 98.5%, and the accuracy rate of identifying asymptomatic infected patients is as high as 100%. Recently, it has been reported that the model has been used to identify Omicron virus.
"ZiDongtai realized the unified expression of graph-text-sound semantics for the first time, and it has the ability of cross-modal understanding and generation." Wang Jinqiao said, "At present, the' All-Media Multimodal Large Model R&D Plan' released by Xinhua News Agency has realized the unified modeling of all-media data understanding and generation, and built a full-stack domestic media artificial intelligence platform, which has been tentatively applied to the textile industry, automobile industry quality inspection and other scenarios."
On February 7th, 65438, official website, the Ministry of Science and Technology issued three letters supporting Harbin, Shenyang and Zhengzhou to build a new generation of national artificial intelligence innovation and development pilot zone. So far, there are 18 national experimental zones for the innovation and development of a new generation of artificial intelligence in China, which will lead the innovation and development of artificial intelligence in China.
"China is promoting the ecological development of artificial intelligence and building a good ecology." Wu Fei said, "At present, there are five national new-generation artificial intelligence development and innovation platforms 15, eight national new-generation artificial intelligence innovation and development experimental zones 18, eight artificial intelligence innovation and application experimental zones, and artificial intelligence undergraduate professional and interdisciplinary talent training carriers established by universities."
"One is a big model, and the other is a combination of artificial intelligence and basic disciplines." Sun Maosong told China Journal of Science that "the basic abilities of language model, graphic model and even multimodal model have been fully demonstrated, and their position as the basic soft facilities of intelligent information processing has been determined. At the same time, it is not simply to expand the scale, but to challenge the integration and computing power of digital resources. Although its limitations are obvious, it has some' strange' properties (such as small sample learning, double decline in depth, task adjustment based on prompts, etc.). ) has aroused scholars' expectations for a super-large parameter scale or a qualitative change, thus laying the foundation for new breakthroughs. "
This year, the field of artificial intelligence has moved from "refining model" to "refining model", from the order of 100 billion to the order of trillion. In the field of big models, there seems to be no biggest, only bigger.
In March, Beijing Zhiyuan Artificial Intelligence Research Institute released China's first ultra-large-scale artificial intelligence model "Enlightenment 1.0". In June, Zhiyuan rewrote his own record and released Enlightenment 2.0, with a parameter scale of 1.75 trillion; In September, Inspur Institute of Artificial Intelligence launched a huge Chinese language model-source 1.0, with parameters of 245.7 billion; 165438+ 10. In October, Aridharma Institute announced the latest progress of the multi-modal large model M6, with the parameters jumping from trillion to 10 trillion; 65438+February, Pengcheng Lab and Baidu jointly released the world's first 100 billion-level knowledge enhancement model-Pengcheng-Baidu Wenxin, with a parameter scale of 260 billion.
Accordingly, recently, Aauto Quicker and the Federal Institute of Technology in Zurich put forward a new recommendation system, Persia, which can support the model training with1000000000000000000000000000-parameter at most.
On the other hand, artificial intelligence is constantly attacking the city in the field of basic disciplines.
In July, the research results of DeepMind's artificial intelligence program Alphafold2 reached the peak of Nature. In the field of structural biology research, artificial intelligence may lead biology, medicine and pharmacy into a new world; 1 1 In June, 2006, researchers at the University of Southern California in the United States used brain-computer connection devices to let monkeys play games and treadmills, thus studying neural activity data; In February 65438, the machine learning framework developed by DeepMind has helped people discover two new conjectures in the field of pure mathematics, and demonstrated the potential of machine learning to support mathematical research.
"This year, the application of artificial intelligence in various industries has also made great achievements." Sun Maosong said, "The combination of artificial intelligence and basic disciplines has shown great potential, and many top papers have been published, which has shown a strong trend, that is,' artificial intelligence+basic science' is promising."
Author Zhang Shuanghu
Brain-computer interface, AR glasses, intelligent voice, EMG bracelet, long-distance gesture recognition ... 202 1 The field of human-computer interaction has surged from basic research to application. Whether it is intelligent health, meta-universe, or the vigorous development of autonomous driving, it seems that human-computer interaction is at the door of industrialization.
"The ultra-flexible neural electrode developed by us in Qualcomm has passed the examination and approval of scientific research and clinical ethics, and the human clinical trial of brain-computer interface will be conducted soon." Hu Tao, deputy director of Shanghai Institute of Microsystems, China Academy of Sciences, and deputy director of State Key Laboratory of Sensing Technology, told China Science News that "collecting large-scale neuron signals of human brain safely and stably, and carrying out closed-loop regulation, will realize the recovery of patients' perception and motor function. "
Brain-computer interface technology brings more and more convenience to patients. In May this year, researchers at Stanford University published a cover paper in Nature, and developed a brain-computer interface system in the cerebral cortex, which can decode the handwritten actions imagined by paralyzed patients from the neural activities in the motor cortex and convert them into texts. With the help of this system, subjects (paralyzed by spinal cord injury) can type nearly 100 characters per minute, and the off-line accuracy rate after automatic error correction is over 99%.
Not long ago, Musk said that he hopes to use Neuralink's microchip device on humans next year. The chip will be used to treat brain diseases and nervous system diseases, such as spinal cord injury and Parkinson's disease. At present, the relevant technology is waiting for the approval of the US Food and Drug Administration.
"In the field of brain-computer interface, considerable technology has been accumulated, which is expected to become a weapon to solve brain diseases." Hu Tao said, "Everyone is seizing the opportunity of clinical application, and technology may be applied next year. It is expected that within two or three years, there will be a unicorn enterprise comparable to Musk Neuralink in China. "
"Human-computer interaction will give birth to a new trillion-dollar market." This judgment of Yan Qun, a famous professor of Fuzhou University, also contains a huge market of the universe.
Some people say that 202 1 is "the first year of the universe", while others think that this is just "new wine in old bottles". But in any case, the metauniverse has become an unavoidable topic in the field of human-computer interaction this year.
"Hyperuniverse is a combination of virtual reality, augmented reality and mixed reality. In fact, there is nothing new." Liu Wei, director of the Laboratory of Human-Computer Interaction and Cognitive Engineering of Beijing University of Posts and Telecommunications, told China Science Journal that "Hyperuniverse is the future development direction of the real world and the virtual world, but there are still some technical problems that have not been solved well."
In the real world, the problem of man-machine interaction and the mixed problem of man-machine environment system have not been well solved. In the human-computer interaction in the real world, whether in the process of input, processing or output, objective data, subjective information and knowledge are still not perfectly integrated.
Liu Wei believes that whether in the real world or the virtual world, the decision-making of people and machines has a process of "fast decision-making" and "slow decision-making". Human decision-making sometimes depends more on logical decision-making and sometimes more on intuitive decision-making. This "mixed decision-making" is constantly changing, and it is difficult to find the law of change. The problem of machine decision-making in this respect has not been solved.
"The hyperuniverse is still in the early stage of painting cakes." Liu Wei said, "Because its underlying mechanism has not been solved-people can't solve the problem of human-computer interaction perfectly in the real world, nor can they solve it in the meta-universe."
Speaking of human-computer interaction, Liu Wei believes that the second problem that cannot be ignored is "complex field".
"This year's Nobel Prize in Physics has also been awarded to supporters of complex system models that predict climate change." Liu Wei said, "Human-computer interaction is also a complex system, which has both repetitive problems and chaotic and cross-disciplinary collaborative problems."
Liu Wei believes that from the point of view of intelligence, complex systems include three important components, one is people, the other is equipment (man-made things) and the third is environment. This is actually a "man-machine ring system" problem in which many things interact, interweave and overlap.
"In human-computer interaction, machines are good at dealing with' complex' problems, and people are good at managing' miscellaneous' things-cross-domain collaboration, balance between things, etc. Because people haven't found the simple operation law of complex things, to solve all the problems of intelligent products and intelligent systems, we must find their combination points, integration points and interaction points from the system of people, machines and environment. Moreover, people must be in a dominant position in this system. "
The third phenomenon that attracted Liu Wei's attention in the field of human-computer interaction was that "artificial intelligence helped mathematicians discover some laws". "Recently, DeepMind has developed a machine learning framework that can help mathematicians discover new conjectures and theorems." Liu Wei said, "Artificial intelligence is a basic mathematical tool, and mathematics also reflects some basic laws. If artificial intelligence can help mathematicians deal with some mathematical problems, then people will better understand the simple laws of complex systems and may make new breakthroughs in human-computer interaction. "
Author Zhang Yunquan (researcher, Institute of Computing Technology, Chinese Academy of Sciences)
This year is a bumper year for supercomputing applications in China.
165438+1At the Global Supercomputing Conference (SC2 1) held in the United States in mid-October, China Supercomputing Application Team won the "Gordon Bell Award", the highest academic award in the field of international high-performance computing applications, for its pioneering simulation of quantum circuits based on the new Shenwei system.
At the same time, in the final of SC 2 1 College Students Supercomputing Competition, Tsinghua University Supercomputing Team won the championship again, achieving four consecutive championships in SC competition. The achievements of these large-scale application software in scalability and performance tuning show that the development of parallel software in China is in the ascendant.
Looking back at the industries driven by supercomputing, we should revisit the term "computing economy". As early as 20 18, we put forward the concept of "computational economy", thinking that the computational economy with supercomputing as the core will become a representative index to measure the development degree of a local digital economy and the main means to convert old and new kinetic energy.
Based on the development trend in recent years, we believe that the current development trend of high-performance computing has fully explained that with the integration and innovation of supercomputing, cloud computing, big data and AI, computing power has become the key to the development of the entire digital information society, and computing power economy has entered the historical stage.
Through the comprehensive analysis of the development status of high-performance computers in China in 20021year, it can be concluded that the current high-performance computing is showing the following characteristics.
First, high-performance computing and cloud computing have been deeply integrated. High performance computing is usually based on MPI, efficient communication, heterogeneous computing and other technologies, and tends to run exclusively. Cloud computing has the ability of flexible deployment and fault tolerance, and supports virtualization, unified resource scheduling and flexible system configuration.
With the development of technology, supercomputing and container cloud are merging and innovating, and high-performance cloud has become a new product and service. AWS, Alibaba Cloud, Tencent, Baidu and the representative of commercial supercomputing "North Long Chaoyun" all launched high-performance cloud services and products based on supercomputing and cloud computing technologies.
Secondly, the application of supercomputing has developed from high precision in the past to a wider and wider direction. With the development of supercomputers, especially the decline in the cost of use, its application fields have also rapidly expanded from the fields of precision research, information security, oil exploration, aerospace and "high cold" scientific calculation with national strategic significance to a wider main battlefield of the national economy, such as pharmaceuticals, gene sequencing, animation rendering, digital movies, data mining, financial analysis and Internet services, which can be said to have penetrated into all walks of life of the national economy.
Judging from the list of China Top 100 in recent years, supercomputing systems used to focus on scientific computing, government, energy, electricity, meteorology and other fields. In the past five years, supercomputing systems deployed by Internet companies have occupied a considerable proportion, and their main applications are cloud computing, machine learning, artificial intelligence, big data analysis and short video. The sharp increase in demand for computing in these fields shows that supercomputing is merging with Internet technology.
Judging from the Linpack performance share of HPC TOP 100 list, computing services occupy the first place with 46%; Supercomputing centers accounted for 24%, ranking second; Artificial intelligence, cloud computing and short video followed closely with 9%, 5% and 4% respectively.
It can be seen that the continuous increase in the proportion of artificial intelligence is closely related to the rapid rise of algorithms and applications such as machine learning and the wide application of deep learning algorithms in big data. Internet companies rediscovered the value of supercomputers, especially heterogeneous supercomputers accelerated by GPU, through deep learning algorithms, and invested heavily in building new systems.
On the whole, computing services, supercomputing centers, artificial intelligence, scientific computing and other fields are the main users of high-performance computing, and the Internet and big data, especially AI, are growing strongly.
Third, the national level has formulated a strategic computing power layout plan. In May this year, the National Development and Reform Commission and other four departments jointly issued the "Implementation Plan for Computing Hub of National Comprehensive Big Data Center Collaborative Innovation System", proposing to build national computing hub nodes in Beijing-Tianjin-Hebei, Yangtze River Delta, Guangdong-Hong Kong-Macao Greater Bay Area, Chengdu-Chongqing, Guizhou, Inner Mongolia, Gansu and Ningxia, starting the implementation of the "Dongdian West Computing" project, urging eastern data to be sent to the west for storage and computing, establishing computing nodes in the west, and improving the uneven layout of digital infrastructure.
Finally, the computing power demand of artificial intelligence has become the main driving force for the development of computing power. Algorithm innovations such as machine learning and deep learning, big data collected through the Internet of Things, sensors, smart phones, smart devices and Internet technologies, and supercomputing power composed of supercomputers and cloud computing are recognized as the "troika" in the era of artificial intelligence, and * * * has set off the latest round of artificial intelligence revolution.
Under the background of the vigorous development of artificial intelligence, virtualized cloud computing has evolved into high-performance container cloud computing, and the integration and innovation of big data, parallel computing and machine learning has become the latest direction of industrial development.
In addition, in terms of intelligent computing evaluation, China has put forward a number of benchmark testing schemes, including AIPerf 500, which is a powerful supplement to the traditional Linpack testing standards.
These developments show that the penetration of supercomputing technology into the industry is accelerating, and we have entered an era of artificial intelligence relying on computing power, which is also one of the inevitable trends of future development. With the increasing demand of users for computing power, computing power economy will surely occupy an important position in the future social development.
Author Wu Yanjun (researcher, Institute of Software, Chinese Academy of Sciences)
The development of open source is remarkable, not just this year. In recent years, many important things have happened in the field of open source.
For example, RISC-V open source instruction set and the rapid rise of its ecology. This is the same as the birth of Linux in the early 1990s. At that time, UNIX and Windows were the mainstream, and few people could have predicted that today's operating system with Linux as the kernel has spread all over people's lives.
At present, more than 80% of the apps people use every day run on the Android operating system with Linux as the kernel, and the operating system running on the back-end server supporting their business is also a Linux distribution.
So today's RISC-V may also be underestimated, thinking that it is immature and difficult to compete with ARM and X86. But perhaps in the future, RISC-V will eventually become the mainstream instruction set ecology in the world like Linux, and its products will cover all aspects.
In 2020 alone, the number of members of RISC-V International (RVI, the new name after the RISC-V Foundation moved to Switzerland) increased by 1.33%. In fact, RVI's move to Switzerland itself is also of great significance. It is a classic case that the open source field keeps its initial intention of not "choosing a side station" in the face of competition from big countries, which is worth learning from other open source foundations around the world.
In China, at the end of 20 19, an open source community of openEuler operating system led by Huawei and participated by Institute of Software of Chinese Academy of Sciences and Kirin Software was formally established. In just two years, the community gathered 7,000 active developers, completed more than 8,000 self-maintenance open source software packages, and gave birth to the commercial distribution of more than 10 vendors.
This is the first real "root community" in the field of basic software in China. Although there is still a gap with Debian and Fedora, which have a history of more than 20 years, an important step has been taken. For academic research, technology research and development, and industrial innovation, there is finally a new platform led by China that can be accumulated for a long time.
At the same time, Huawei launched the HarmonyOS operating system HarmonyOS system after the overseas supply of the Android operating system GMS (Google Mobile Service) was cut off, and launched the open source project OpenHarmony under the Open Atomic Open Source Foundation.
At present, OpenHarmony has attracted many domestic manufacturers to participate in a short time, which also reflects the strong demand of the domestic industry for a new generation of Internet of Everything operating system. Although it still lags behind Android in terms of ecological scale and technical integrity, it has taken the first step to build its own ecology.
This is equivalent to drawing a boundary for the reasonable use of source code, that is, the reasonable use is limited to the interface, and once it goes deep into the implementation code of the interface, it needs to comply with the relevant license. This has important reference significance for the legal definition of open source intellectual property rights.
In May of this year, China Blue Book 202 1 was released. It not only systematically sorts out the current situation of open source talents, projects, communities, organizations, education and business in China, but also gives development suggestions, which also provides reference for the relevant administrative departments of the national government to formulate open source policies and lay out open source strategies, and provides more case references and data support for scientific research institutes, technology enterprises and open source practitioners.
Whether open source software is developing towards the open source software and hardware ecology around the open instruction set, or open source has strict legal boundary constraints, or domestic leading enterprises are trying to solve the "stuck neck" problem through open source exploration, and have achieved certain results ... Many cases point to one direction-the open source trend is unstoppable. Because it stems from the nature of human beings to share knowledge and create collaboratively, it is also an important mode of human civilization handed down from generation to generation in the digital age.
Of course, it is undeniable that there are still many problems in open source, such as the security of open source software supply chain. The security here not only has the problems of software quality and security vulnerabilities in the traditional sense, but also has the problem that open source software cannot be maintained continuously and effectively (for example, OpenSSL has only two part-time maintainers when HeartBleed has problems, and log4j has only three part-time maintainers when it has problems), and there is also the problem of "supply failure" caused by competition among big countries (for example, GitHub once restricted the access of Iranian developers).
With the concentration of open source software on commercial platforms such as GitHub, this problem will become more prominent and even become a big risk. Open source software, an intellectual asset that should belong to all mankind, may become a sharp weapon to implement "long-arm jurisdiction". In order to avoid this problem, public infrastructure such as open source code hosting platform and open source software construction and publishing platform need to be decentralized. The world needs multiple open source software infrastructures to eliminate the threat of political forces to the open source community to the greatest extent.
For China, with open source software becoming an important supporting part of many major infrastructures such as scientific research and industry, open source software itself should also have infrastructure with functions such as code hosting, compilation, construction, testing, release, operation and maintenance, so as to ensure the security and continuity of open source software supply and further enhance the confidence of all walks of life in using open source software.
In the future, core technology innovation and open source contribution will become the new impetus for the development of domestic enterprises, or push China's open source cause to another climax.