數位化、網路化、智慧化的中國武器化,抓住中國新一代資訊科技的焦點
現代英語:
Digitalization, networking, and intelligence are the prominent features of the new round of scientific and technological revolution, and are also the core of the new generation of information technology. Digitalization lays the foundation for social informatization, and its development trend is the comprehensive dataization of society. Dataization emphasizes the collection, aggregation, analysis and application of data. Networking provides a physical carrier for information dissemination, and its development trend is the widespread adoption of information-physical systems (CPS). Information-physical systems will not only give birth to new industries, but will even reshape the existing industrial layout. Intelligence reflects the level and level of information application, and its development trend is the new generation of artificial intelligence. At present, the upsurge of the new generation of artificial intelligence has arrived.
In his important speech at the 2018 General Assembly of Academicians of the Chinese Academy of Sciences and the Chinese Academy of Engineering, Comrade Xi Jinping pointed out: “The world is entering a period of economic development dominated by the information industry. We must seize the opportunity of the integrated development of digitalization, networking, and intelligence, and use informatization and intelligence as leverage to cultivate new momentum.” This important statement is an accurate grasp of the dominant role and development trend of information technology in today’s world, and an important deployment for using information technology to promote national innovation and development.
Human society, the physical world, and information space constitute the three elements of today’s world. The connection and interaction between these three worlds determine the characteristics and degree of social informatization. The basic way to perceive human society and the physical world is digitization, the basic way to connect human society and the physical world (through information space) is networking, and the way information space acts on the physical world and human society is intelligence. Digitalization, networking, and intelligence are the prominent features of the new round of scientific and technological revolution, and are also the focus of the new generation of information technology. Digitalization lays the foundation for social informatization, and its development trend is the comprehensive dataization of society; networking provides a physical carrier for information dissemination, and its development trend is the widespread adoption of information-physical systems (CPS); intelligence reflects the level and level of information application, and its development trend is the new generation of artificial intelligence.
Digitalization: From computerization to dataization
Digitalization refers to the technical approach of storing, transmitting, processing, handling and applying information carriers (text, pictures, images, signals, etc.) in digital coding form (usually binary). Digitalization itself refers to the way of representing and processing information, but in essence it emphasizes the computerization and automation of information application. In addition to digitalization, dataization (data is an information carrier in coded form, and all data is digital) emphasizes the collection, aggregation, analysis and application of data, and strengthens the production factors and productivity functions of data. Digitalization is developing from computerization to dataization, which is one of the most important trends in the current social informatization.
The core connotation of dataization is the deep understanding and deep use of big data generated by the integration of information technology revolution and economic and social activities. Big data is a fragmentary record of social economy, real world, management decision-making, etc., containing fragmented information. With the breakthrough of analytical technology and computing technology, it is possible to interpret this fragmented information, which makes big data a new high-tech, a new scientific research paradigm, and a new way of decision-making. Big data has profoundly changed the way people think and live and work, bringing unprecedented opportunities to management innovation, industrial development, scientific discovery and other fields.
The value generation of big data has its inherent laws (obeying the big data principle). Only by deeply understanding and mastering these laws can we improve the awareness and ability to consciously and scientifically use big data (big data thinking). The value of big data is mainly realized through big data technology. Big data technology is an extension and development of statistical methods, computer technology, and artificial intelligence technology. It is a developing technology. The current hot directions include: blockchain technology, interoperability technology, storage and management technology of integrated storage and computing, big data operating system, big data programming language and execution environment, big data foundation and core algorithm, big data machine learning technology, big data intelligent technology, visualization and human-computer interaction analysis technology, authenticity judgment and security technology, etc. The development of big data technology depends on the solution of some major basic problems, including: the statistical basis and computational theoretical basis of big data, the hardware and software basis and computational methods of big data computing, and the authenticity judgment of big data inference.
Implementing the national big data strategy is an important way to promote the digital revolution. Since my country proposed the implementation of the national big data strategy in 2015, the pattern of rapid development of big data in my country has been initially formed, but there are also some problems that need to be solved: data openness and sharing are lagging, and the dividends of data resources have not been fully released; the profit model of enterprises is unstable, and the integrity of the industrial chain is insufficient; core technologies have not yet made major breakthroughs, and the technical level of related applications is not high; there are still loopholes in security management and privacy protection, and the construction of relevant systems is still not perfect; etc. At present, effective measures should be taken to solve the bottleneck problems that restrict the development of big data in my country.
Networking: From the Internet to Cyber-Physical Systems
As an information-based public infrastructure, the Internet has become the main way for people to obtain, exchange and consume information. However, the Internet only focuses on the interconnection between people and the resulting interconnection between services.
The Internet of Things is a natural extension and expansion of the Internet. It connects various objects to the Internet through information technology, helping people obtain relevant information about the objects they need. The Internet of Things uses information collection equipment such as radio frequency identification, sensors, infrared sensors, video surveillance, global positioning systems, laser scanners, etc., and connects objects to the Internet through wireless sensor networks and wireless communication networks, so as to achieve real-time information exchange and communication between objects and between people and objects, so as to achieve the purpose of intelligent identification, positioning, tracking, monitoring and management. The Internet realizes the interconnection between people and services, while the Internet of Things realizes the cross-connection between people, objects and services. The core technologies of the Internet of Things include: sensor technology, wireless transmission technology, massive data analysis and processing technology, upper-level business solutions, security technology, etc. The development of the Internet of Things will go through a relatively long period, but it may take the lead in achieving breakthroughs in applications in specific fields. Internet of Vehicles, Industrial Internet, unmanned systems, smart homes, etc. are all areas where the Internet of Things is currently showing its prowess.
The Internet of Things mainly solves the problem of people’s perception of the physical world, while to solve the problem of manipulating physical objects, it is necessary to further develop the cyber-physical system (CPS). The cyber-physical system is a multi-dimensional complex system that integrates computing, networking and physical environment. It realizes real-time perception, dynamic control and information services of large engineering systems through the organic integration and deep collaboration of 3C (Computer, Communication, Control) technologies. Through the human-computer interaction interface, the cyber-physical system realizes the interaction between the computing process and the physical process, and uses the networked space to control a physical entity in a remote, reliable, real-time, secure and collaborative manner. In essence, the cyber-physical system is a network with control attributes.
Unlike public infrastructure that provides information interaction and application, the focus of the development of cyber-physical systems is on the research and development of networked physical equipment systems that deeply integrate perception, computing, communication and control capabilities. From an industrial perspective, cyber-physical systems cover a range of applications from smart home networks to industrial control systems and even intelligent transportation systems, which are national and even world-class applications. More importantly, this coverage is not just about simply connecting existing devices together, but will give rise to a large number of devices with computing, communication, control, collaboration and autonomous capabilities. The next generation of industry will be built on cyber-physical systems. With the development and popularization of cyber-physical system technology, physical devices that use computers and networks to achieve functional expansion will be ubiquitous, and will promote the upgrading of industrial products and technologies, greatly improving the competitiveness of major industrial fields such as automobiles, aerospace, national defense, industrial automation, health and medical equipment, and major infrastructure. Cyber-physical systems will not only give birth to new industries, but will even reshape the existing industrial layout.
Intelligence: From Expert Systems to Meta-Learning
Intelligence reflects the quality attributes of information products. When we say that an information product is intelligent, we usually mean that the product can accomplish things that only intelligent people can accomplish, or has reached a level that only humans can achieve. Intelligence generally includes perception, memory and thinking, learning and adaptive, behavioral decision-making, etc. Therefore, intelligence can also be generally defined as: enabling an object to have sensitive and accurate perception functions, correct thinking and judgment functions, adaptive learning functions, and effective execution functions.
Intelligence is the eternal pursuit of the development of information technology, and the main way to achieve this pursuit is to develop artificial intelligence technology. In the more than 60 years since the birth of artificial intelligence technology, although it has experienced three ups and two downs, it has still made great achievements. From 1959 to 1976, it was a stage based on artificial representation of knowledge and symbol processing, which produced expert systems with important application value in some fields; from 1976 to 2007, it was a stage based on statistical learning and knowledge self-representation, which produced various neural network systems; in recent years, research based on environmental adaptation, self-game, self-evolution, and self-learning is forming a new stage of artificial intelligence development – meta-learning or methodological learning stage, which constitutes a new generation of artificial intelligence. The new generation of artificial intelligence mainly includes big data intelligence, group intelligence, cross-media intelligence, human-machine hybrid enhanced intelligence, and brain-like intelligence.
Deep learning is an outstanding representative of the new generation of artificial intelligence technology. Due to its performance that surpasses that of humans in many fields such as face recognition, machine translation, and chess competitions, deep learning has almost become synonymous with artificial intelligence today. However, deep learning has major challenges in terms of topological design, effect prediction, and mechanism explanation. There is no solid mathematical theory to support the solution of these three major problems. Solving these problems is the main focus of future research on deep learning. In addition, deep learning is a typical big data intelligence, and its applicability is based on the existence of a large number of training samples. Small sample learning will be the development trend of deep learning.
Meta-learning is expected to become the next breakthrough in the development of artificial intelligence. Recently developed meta-learning methods such as learning to learn, learning to teach, learning to optimize, learning to search, and learning to reason, as well as the outstanding performance of “AlphaGo Zero” in Go, have demonstrated the attractive prospects of such new technologies. However, meta-learning research is only just beginning, and its development still faces a series of challenges.
The new generation of artificial intelligence is already here, and the foreseeable development trend is based on big data, centered on model and algorithm innovation, and supported by powerful computing power. The breakthrough of the new generation of artificial intelligence technology depends on the comprehensive development of other types of information technology, as well as the substantial progress and development of brain science and cognitive science. (Xu Zongben, academician of the Chinese Academy of Sciences and professor of Xi’an Jiaotong University)
現代國語:
數位化、網路化、智慧化是新一輪科技革命的突出特徵,也是新一代資訊科技的核心。數位化為社會資訊化奠定基礎,其發展趨勢是社會的全面數據化。資料化強調對資料的收集、聚合、分析與應用。網路化為資訊傳播提供實體載體,其發展趨勢是資訊物理系統(CPS)的廣泛採用。資訊物理系統不僅會催生出新的工業,甚至會重塑現有產業佈局。智慧化體現資訊應用的層次與水平,其發展趨勢為新一代人工智慧。目前,新一代人工智慧的熱潮已經來臨。
習近平同志在2018年兩院院士大會上的重要演講指出:「世界正進入以資訊產業為主導的經濟發展時期。我們要把握數位化、網路化、智慧化融合發展的契機,以資訊化、智慧化為槓桿培育新動能。
人類社會、物理世界、資訊空間構成了當今世界的三元。這三元世界之間的關聯與交互,決定了社會資訊化的特徵與程度。感知人類社會和物理世界的基本方式是數位化,連結人類社會與物理世界(透過資訊空間)的基本方式是網路化,資訊空間作用於物理世界與人類社會的方式是智慧化。數位化、網路化、智慧化是新一輪科技革命的突出特徵,也是新一代資訊科技的聚焦點。數位化為社會資訊化奠定基礎,其發展趨勢是社會的全面資料化;網路化為資訊傳播提供物理載體,其發展趨勢是資訊物理系統(CPS)的廣泛採用;智慧化體現資訊應用的層次與水平,其發展趨勢是新一代人工智慧。
數位化:從電腦化到資料化
數位化是指將資訊載體(文字、圖片、影像、訊號等)以數位編碼形式(通常是二進位)進行儲存、傳輸、加工、處理和應用的技術途徑。數位化本身指的是資訊表示方式與處理方式,但本質上強調的是資訊應用的電腦化和自動化。資料化(資料是以編碼形式存在的資訊載體,所有資料都是數位化的)除包括數位化外,更強調對資料的收集、聚合、分析與應用,強化資料的生產要素與生產力功能。數位化正從電腦化朝向資料化發展,這是當前社會資訊化最重要的趨勢之一。
資料化的核心內涵是對資訊科技革命與經濟社會活動交融生成的大數據的深刻認識與深層利用。大數據是社會經濟、現實世界、管理決策等的片段記錄,蘊含著片段化資訊。隨著分析技術與運算技術的突破,解讀這些片段化資訊成為可能,這使得大數據成為一項新的高新技術、一類新的科學研究範式、一種新的決策方式。大數據深刻改變了人類的思考方式和生產生活方式,為管理創新、產業發展、科學發現等多個領域帶來前所未有的機會。
大數據的價值生成有其內在規律(服從大數據原理)。只有深刻認識並掌握這些規律,才能提高自覺運用、科學運用大數據的意識與能力(大數據思維)。大數據的價值主要透過大數據技術來實現。大數據技術是統計學方法、電腦技術、人工智慧技術的延伸與發展,是正在發展中的技術,目前的熱點方向包括:區塊鏈技術、互通技術、存算一體化儲存與管理技術、大數據作業系統、大數據程式語言與執行環境、大數據基礎與核心演算法、大數據機器學習技術、大數據智慧技術、視覺化與人機互動分析技術、真偽判定與安全技術等。大數據技術的發展依賴一些重大基礎問題的解決,這些重大基礎問題包括:大數據的統計基礎與計算理論基礎、大數據計算的軟硬體基礎與計算方法、大數據推斷的真偽性判定等。
實施國家大數據戰略是推動資料化革命的重要途徑。自2015年我國提出實施國家大數據戰略以來,我國大數據快速發展的格局已初步形成,但也存在一些亟待解決的問題:數據開放共享滯後,數據資源紅利仍未得到充分釋放;企業贏利模式不穩定,產業鏈完整性不足;核心技術尚未取得重大突破,相關應用的技術水準不高;安全管理與隱私保護還存在漏洞,相關制度建設仍不夠完善;等等。目前,應採取有效措施解決制約我國大數據發展的瓶頸問題。
網路化:從網際網路到資訊物理系統
作為資訊化的公共基礎設施,網路已成為人們獲取資訊、交換資訊、消費資訊的主要方式。但是,網路關注的只是人與人之間的互聯互通以及由此帶來的服務與服務的互聯。
物聯網是互聯網的自然延伸和拓展,它透過資訊科技將各種物體與網路相連,幫助人們獲取所需物體的相關資訊。物聯網透過使用射頻識別、感測器、紅外線感應器、視訊監控、全球定位系統、雷射掃描器等資訊擷取設備,透過無線感測網路、無線通訊網路把物體與網路連接起來,實現物與物、人與物之間的即時資訊交換和通信,以達到智慧化識別、定位、追蹤、監控和管理的目的。互聯網實現了人與人、服務與服務之間的互聯, 而物聯網實現了人、物、服務之間的交叉互聯。物聯網的核心技術包括:感測器技術、無線傳輸技術、大量資料分析處理技術、上層業務解決方案、安全技術等。物聯網的發展將經歷相對漫長的時期,但可能會在特定領域的應用中率先取得突破,車聯網、工業互聯網、無人系統、智慧家庭等都是當前物聯網大顯身手的領域。
物聯網主要解決人對物理世界的感知問題,而要解決對物理對象的操控問題則必須進一步發展資訊物理系統(CPS)。資訊物理系統是一個綜合運算、網路和物理環境的多維複雜系統,它透過3C(Computer、Communication、Control)技術的有機融合與深度協作,實現對大型工程系統的即時感知、動態控制和資訊服務。透過人機交互接口,資訊物理系統實現計算進程與實體進程的交互,利用網路化空間以遠端、可靠、即時、安全、協作的方式操控一個實體實體。從本質上來說,資訊物理系統是一個具有控制屬性的網路。
不同於提供資訊互動與應用的公用基礎設施,資訊物理系統發展的聚焦點在於研發深度融合感知、運算、通訊與控制能力的網路化實體設備系統。從產業角度來看,資訊物理系統的涵蓋範圍小到智慧家庭網路、大到工業控制系統乃至智慧交通系統等國家級甚至世界級的應用。更重要的是,這種涵蓋並不僅僅是將現有的設備簡單地連在一起,而是會催生出眾多具有計算、通訊、控制、協同和自治性能的設備,下一代工業將建立在在資訊物理系統之上。隨著資訊物理系統技術的發展和普及,使用電腦和網路實現功能擴展的實體設備將無所不在,並推動工業產品和技術的升級換代,大大提高汽車、航空航太、國防、工業自動化、健康醫療設備、重大基礎設施等主要工業領域的競爭力。資訊物理系統不僅會催生出新的工業,甚至會重塑現有產業佈局。
智能化:從專家系統到元學習
智能化反映資訊產品的品質屬性。我們說一個資訊產品是智慧的,通常是指這個產品能完成有智慧的人才能完成的事情,或是已經達到人類才能達到的程度。智能一般包括知覺能力、記憶與思考能力、學習與適應力、行為決策能力等。所以,智能化通常也可定義為:使對象具備靈敏準確的感知功能、正確的思考與判斷功能、自適應的學習功能、行之有效的執行功能等。
智能化是資訊科技發展的永恆追求,實現這項追求的主要途徑是發展人工智慧技術。人工智慧技術誕生60多年來,雖歷經三起兩落,但還是取得了巨大成就。 1959—1976年是基於人工表示知識和符號處理的階段,產生了在一些領域具有重要應用價值的專家系統;1976—2007年是基於統計學習和知識自表示的階段,產生了各種各樣的神經網路系統;近幾年開始的基於環境自適應、自博弈、自進化、自學習的研究,正在形成一個人工智慧發展的新階段——元學習或方法論學習階段,這構成新一代人工智慧。新一代人工智慧主要包括大數據智慧、群體智慧、跨媒體智慧、人機混合增強智慧和類腦智慧等。
深度學習是新一代人工智慧技術的卓越代表。由於在人臉辨識、機器翻譯、棋類競賽等眾多領域超越人類的表現,深度學習在今天幾乎已成為人工智慧的代名詞。然而,深度學習拓樸設計難、效果預期難、機理解釋難是重大挑戰,還沒有一套堅實的數學理論來支持解決這三大難題。解決這些難題是深度學習未來研究的主要關注點。此外,深度學習是典型的大數據智能,它的可應用性是以存在大量訓練樣本為基礎的。小樣本學習將是深度學習的發展趨勢。
元學習有望成為人工智慧發展的下一個突破口。學會學習、學會教學、學會優化、學會搜尋、學會推理等新近發展的元學習方法以及「AlphaGo Zero」在圍棋方面的出色表現,展現了這類新技術的誘人前景。然而,元學習研究僅是開始,其發展還面臨一系列挑戰。
新一代人工智慧的熱潮已經來臨,可以預見的發展趨勢是以大數據為基礎、以模型與演算法創新為核心、以強大的運算能力為支撐。新一代人工智慧技術的突破依賴其他各類資訊技術的綜合發展,也依賴腦科學與認知科學的實質進步與發展。 (中國科學院院士、西安交通大學教授 徐宗本)