In 2020, the epidemic will shake the world and the economy will fluctuate violently. Both countries and enterprises hope to grasp the "pulse" of future technology, dispel the shadow of economic downturn, and realize organizational upgrading and iteration.
In the new decade, what key technologies will reshape our society and change our lives?
Recently, the Technical Committee of JD.com released the " Technology Reconstructing Social Supply Chain-White Paper on Future Technology Trends " ( PDF download is available at the end of the article ), which fully demonstrated to the industry that JD.com's digital intelligent social supply chain has the advantages of artificial intelligence, Internet of Things, and blockchain. Research on technology trends in five key technology areas, including autonomous systems, and next-generation technologies, as well as technological empowerment achievements in five major scenarios including retail, health, logistics, finance, and cities.
Among them, artificial intelligence, Internet of Things, blockchain, autonomous systems, and next-generation computing are becoming the main keywords of future technological trends, and actively integrate these five technologies with business scenarios:
- Artificial intelligence (including machine learning, deep learning, and federated learning, etc.) will make important breakthroughs in basic theories and key technologies in the fields of machine learning, machine vision, and natural language processing, thereby enhancing the trustworthiness of AI.
- The development of IoT devices, 5G communication technology, and edge computing technology will enable the Internet of Things to achieve leapfrog popularity, and the digital twin world will become possible.
- The consensus mechanism, distributed storage, and security protection of the blockchain play an important role in ensuring the credible and reliable dissemination of data. The development of blockchain will build a new model of digital intelligent supply chain based on the business model of mutual trust and cooperation, and form a new social ecology of complete trust between people, sharing and co-construction.
- Autonomous systems (autonomous driving, high-precision maps, autonomous environment perception, etc.) will enter the era of advanced automation and autonomy. Human brains, cloud brains, and terminal brains will merge and coexist with biological, digital, and mechanical bodies to achieve flexible and intelligent integration of humans and machines.
- Next-generation computing (including quantum computing, cloud computing, heterogeneous computing, etc.) has the characteristics of multi-scenario, security, and cloud-edge-device collaboration and integration, and will become an important infrastructure for a digitally intelligent society.
The following is an extended interpretation of "Technology Reconstructing the Social Supply Chain-White Paper on Future Technology Trends".
Trend 1: The three key technologies of "artificial intelligence" are moving towards the "deep water area"
Regarding the future development direction of artificial intelligence, in this white paper, three mainstream technologies are mainly analyzed:
- machine learning
- Machine Vision (Vision)
- Natural language processing (listening, reading, speaking)
At present, artificial intelligence technology shows more "artificial" intelligence, relying on artificial intelligence to help machines broaden the scope of learning.
The technical committee of JD.com believes that although the current method has not yet endowed the machine with complete cognitive intelligence or decision-making intelligence, it has already generated commercial value in multiple industries and application scenarios such as online retail, financial risk control, and smart cities.
One of the fastest growing branches in the field of machine learning is deep learning based on neural networks.
At present, engineering projects such as big data, unmanned vehicles, computer vision, and intelligent supply chains in the industry all use deep learning algorithms. Among them, for deep learning with large amounts of data, model training is a key step, and it is also time-consuming and consumes GPU resources, which seriously limits the advancement of various deep learning applications.
For this pain point, JD.com proposed a novel deep learning distributed optimization theory and algorithm—Features replay algorithm (Features replay), and verified through experiments that the feature replay algorithm is better than the traditional backpropagation algorithm without performance loss. (Backpropagation), saving at least 34.4% of the training time in terms of speed.
In addition, with the continuous emergence of new scenario problems, machine learning paradigms combined with specific scenarios are constantly being proposed, representing the direction of future machine learning development, such as federated learning, migration learning, confrontational learning, distributed learning, meta-learning, scalable Explanatory Learning, Lightweight Learning, Quantum Machine Learning, and more.
And JD Technology is also actively following up on these machine learning development directions. For example, in October 2020, it officially launched its self-developed federated learning platform - Fedlearn. At present, JD Technology's gold bars and baitiao business models have fully realized the upgrade from traditional modeling to federated modeling, and the risk identification ability and traffic conversion rate of the models under the federated model have been improved. In addition to its application in risk control, the Fedlearn platform is also widely used in smart cities and other fields.
In addition to machine learning, the white paper also focuses on the analysis of two mainstream technologies, machine vision and natural language processing.
The Jingdong Group Technical Committee believes that in the field of machine vision, the accuracy of large-scale visual analysis and recognition has been continuously improved, and it has landed in some practical application scenarios, but the application of complex open scenarios has not yet been widely spread. At present, machine vision technology is rapidly popularized in scenarios such as security, industrial inspection, and automatic driving. It can be specifically used in mobile phones, driving recorders, and security cameras.
In terms of natural language processing, the concept of "familiar with the three hundred Tang poems, and can sing without writing poems" is not only applicable to humans, but also applicable to machines.
In the future, scenarios such as intelligent customer service, intelligent marketing, travel consultation, and medical consultation can provide anthropomorphic services close to the level of industry experts with the accumulation and use of a large amount of data. It is reported that on the day of "JD.com 11.11" in 2020, JD.com's emotional intelligent customer service provided a total of 65.52 million times, and JD.com's intelligent analysis platform was called more than 100 million times.
Trend 2: "Internet of Things" application scenarios will usher in a big explosion
The Internet of Things (IoT, The Internet of things) is an Internet in which people, things, and services are ubiquitously connected.
The continuous innovation of IoT technology is closely integrated with social consumption and industrial production, gradually forming new information industries such as consumer Internet of Things (C-end), industrial Internet of Things (B-end), and digital cities.
Taking the consumer Internet of Things as an example, the scenarios of the consumer Internet of Things include smart homes and wearable devices. ) and other scenes in families and even larger places are opportunities for the development of smart homes.
The development goal of the Internet of Things industry is to become a connection hub between the physical world and the digital world, integrate the cultural and world views of the two worlds, improve the life experience of human society, and enhance industrial efficiency. According to statistics from iResearch, as of 2019, the number of IoT device connections in China has reached 4.57 billion, and it is expected that the number of IoT connections in China will continue to grow at a compound annual growth rate of 27.8% from 2019 to 2025.
At the same time, according to the statistics of the Ministry of Industry and Information Technology of China, as of June 2020, the number of end users of the cellular Internet of Things has reached 1.106 billion, of which end users used in smart manufacturing, smart transportation, and smart public utilities accounted for 21.1% respectively. , 18.2%, 21.4%. China's Internet of Things is growing rapidly in terms of equipment access, and its application scenarios are flourishing.
The Internet of Things technology has a wealth of application scenarios in JD.com's business system. JD.com focuses on the development of the Internet of Things and related technologies to improve the operational efficiency of its own logistics, retail, health, finance, urban and other businesses.
Since 2014, JD.com has gradually opened up its IoT technology to the outside world to empower various intelligent fields. At present, JD.com has formed industry solutions including smart home, smart logistics, smart finance, smart medical care, industrial IoT, smart real estate, smart car and other fields.
Compared with other manufacturers, the advantages of self-operated e-commerce attributes enable JD.com to integrate smart home products across brands and categories, forming more mature digital and scene twinning capabilities. Relying on the cloud-native technology of JD.com, JD.com has formed a complete single-product intelligent solution, providing a reliable cloud service platform, cost-effective modules, and APP-side control methods for smart home products. It can support more than 10 kinds of intelligent protocols, and can Quickly help brand owners realize product intelligence.
In addition, JD.com also provides a smart assistant empowerment solution, which enables products to have hearing, vision and understanding capabilities, and realizes functions such as playing music, shopping, and controlling home equipment, making products more intelligent. At present, Jingdong Smart Home Jingyuzuo ecological brand has become a brand with open device control capabilities in the field of smart home, providing services for more than 230 categories, more than 1,000 brands, and more than 4,000 products, connecting more than 200 million devices .
The popularization of AR/VR/MR technology can enable people to change the operation of real equipment through remote control of virtual equipment; the development of brain-computer interface can allow equipment to directly sense people's needs and respond in a timely manner, further simplifying the operation process of equipment; People can use the above technologies to change the operation of real equipment by operating digital twins, which greatly improves the convenience of life and work.
The development of Internet of Things technology can also allow manufacturers on the product supply side to directly perceive the product use needs of consumers on the demand side, mobilize their own products to serve consumers, and jump out of the traditional business model that consumers need to "own first and enjoy later". .
Manufacturers will transition from the current model of selling products to the model of selling products + services, and then develop to the final model of selling services. Automobile manufacturers can develop from the current model of selling cars + after-sales maintenance services to a travel service model; aero-engine manufacturers (such as Rolls-Royce) can develop from selling engines + maintenance services to a rental service model, eliminating the gap between product manufacturers and consumers. It is a rigid requirement that the operator needs to connect the two parties through physical equipment.
Trend 3: Consortium chain has become the mainstream form of enterprise application "block chain"
"Blockchain" is a general term for a combination of technologies. It is an information technology term born after combining theories and technologies such as mathematics, cryptography, Internet communication, distributed storage, and consensus protocols. It supports the shared storage and storage of digital information. Joint authentication can reach a common resolution through multi-party voting in the fields of justice, trade, finance, etc., and realize collaboration, trust and concerted action among multiple subjects.
At present, blockchain technology is still in the stage of rapid development and iteration. Under the guidance of relevant policies such as "digital economy" and "new infrastructure", blockchain technology is transforming from small-scale application exploration to large-scale application practice. In addition, with the injection of more resources and the mining of real needs, blockchain technology is accelerating the transformation of all walks of life. The independent innovation of blockchain core technology and the self-development of the underlying architecture system have become the basis for its development. trend. In the context of the country's claim of independent intellectual property rights, the blockchain has risen to the national strategy and is expected to become a new engine of the digital economy.
Smart contracts are codes that the blockchain system can automatically execute according to preset conditions, and are the core components of blockchain 2.0. Using smart contracts, users can develop various decentralized applications. Today, the emergence of various blockchain applications such as e-wallets, games, exchanges, markets, and voting systems all benefit from smart contracts.
In the future, blockchain 3.0 is based on smart contracts and deeply embedded in various blockchain applications in social life. People can use smart contracts for financial management, entertainment, learning, etc.
Blockchains are divided into public chains, private chains, and alliance chains. At present, the alliance chain has become the mainstream form of enterprise application blockchain technology.
Under the single-chain architecture, the existing blockchain technology has bottlenecks in performance, capacity, privacy, isolation, scalability, etc., which drives the rise of the alliance chain.
The alliance chain realizes multi-centralization and supports cross-departmental business collaboration through a consensus mechanism for alliance members. Relying on alliance chain technology and alliance collaboration to establish a sustainable, safe and credible support framework, to achieve various forms of chain communication between the underlying platform and application requirements.
In the current blockchain implementation scenario, the main business is the alliance chain formed between enterprises. With the development of business, each alliance chain data island will be formed.
In the future, all alliances will be connected to each other. Relying on industrial-grade cross-chain interoperability technology, the intercommunication and interoperability of heterogeneous cloud environments and heterogeneous blockchain bottom layers can be realized, cross-chain integration can be realized, and large-scale heterogeneous chain cross-chain can be supported. synergy.
Trend 4: "Autonomous system" advanced automation, automatic driving is being implemented in batches
An autonomous system refers to a control system with certain autonomy and autonomy, combined with technologies such as artificial intelligence and the Internet of Things, and realizes system autonomy through real-time control and decision-making.
At present, autonomous systems are mainly used in areas such as unmanned vehicles, drones, and mobile robots. Autonomous systems can replace or assist humans to complete tasks in specific environments, greatly improving the range of human perception and expanding human behavioral capabilities.
In the military and civilian fields, autonomous systems can meet the automation and intelligence requirements of multiple scenarios. Through the organic collaboration of the entire integrated system, system performance can be greatly improved, thereby meeting the requirements for overall efficiency and production capacity.
Autopilot technology is an important part of autonomous systems, designed to operate motor vehicles automatically and safely without any active human intervention.
At present, with the gradual maturity of the industrial chain, autonomous driving is being implemented in batches. Autonomous driving companies and vehicle production companies form an upstream and downstream relationship in the industrial chain. Vehicle production companies focus on reducing vehicle production costs and improving their stability and consistency, while autonomous driving companies focus on providing high-performance autonomous driving services to transform them into Products are provided to customers. The popularity of autonomous driving has activated the relevant smart component market, and computing units, chips, sensors, high-precision maps, and simulation cloud platform technologies will develop rapidly, promoting rapid upgrades and iterations of autonomous driving technology.
JD.com’s unmanned delivery vehicle has undergone several iterations since its birth in 2016. It is currently equipped with an L4-level autonomous driving chip, and has a data simulation platform, robot management platform, and industrial control platform, which can realize autonomous environment perception, high-precision positioning, and planning and prediction. Function.
Trend 5: "Next Generation Computing" such as cloud computing and quantum computing will usher in new development
In recent years, with the continuous growth of data scale, artificial intelligence, especially deep learning algorithms, have been applied in more and more industries, which put forward higher and higher requirements and challenges for computing systems. Driven by new technologies, new scenarios, and new challenges, the computing field is undergoing a profound transformation.
In the next generation of computing, computing tasks are gradually moving from the user end to the edge end and remote cloud server with more powerful computing capabilities, and are developing towards a hierarchical structure of cloud, edge, and end collaborative computing.
Cloud computing hardware has become the main battlefield for computing technology innovation. A variety of new computing architectures and processors have been developed and put into use. In a few edge applications, such as autonomous driving and security, dedicated computing chips have been approved and put into use, and there is still a vast application space waiting to be covered by new computing technologies.
Therefore, it is difficult to have a general-purpose computing hardware suitable for computing needs of all edges.
The integrated circuit process is gradually approaching the physical limit of the transistor diameter after breaking through 5nm. Although the current integrated circuit process continues to move towards 1nm, its manufacturing cost has increased significantly.
Whether the exponential growth of computing power can be maintained at the same cost level may become the key to the advancement of next-generation computing. In the attempt to find alternative computing power of integrated circuit technology, quantum computing is considered to be the most likely technology to supplement the computing power of traditional integrated circuits.
In a calculation case in the experiment, it takes 10,000 years to calculate with the most advanced supercomputer, but the quantum computer only takes 200 seconds. And China's "Jiuzhang" quantum computing prototype achieves 76 qubits.
my country's quantum computing has gradually stepped out of the laboratory. Quantum computing has the ability to accelerate exponentially on specific computing tasks. It is expected to become a new form of computing in the "post-Moore's Law" era and form a benign incentive for potential commercial applications.
The white paper points out that the cloud platform has become an important way for quantum computing application exploration and industrial promotion. Many domestic and foreign technology companies and start-ups, including Benyuan Quantum and Shenzhen Liangxuan, have successively launched their own quantum computing cloud platforms and services.
Climb high and look far, walk on the road
Open your eyes, and some technologies such as machine vision, natural language processing, 5G communication, and autonomous driving will iterate rapidly, some will be implemented on a large scale, and some will gradually mature;
Ascending the heights and looking far, quantum computing, AR/VR, brain-computer interface and other technological applications are widely popularized. Although it is still extremely difficult to complete within three to five years, the timeline is stretched and suddenly enlightened.
At the end of 2020, the state issued the "14th Five-Year Plan" and formulated China's long-term goals for 2035. The "14th Five-Year Plan" emphasizes the need to accelerate digital development and develop strategic emerging industries, and even mentions technological keywords such as big data, artificial intelligence, and quantum information.
Although the epidemic is turbulent, the strategy of rejuvenating the country through science and technology remains unchanged. The picture scroll of 2021 has slowly unfolded. Since people have chosen a distant place, they are walking on the road and have their own future.
