1. Artificial intelligence evolves from perceptual intelligence to cognitive intelligence
Artificial intelligence has reached or surpassed humans in the areas of perceptual intelligence such as speech to text, natural language processing, video understanding, etc; but in the field of cognitive intelligence that requires external knowledge, logical reasoning, or domain migration, it is still in its infancy. Cognitive intelligence will draw inspiration from cognitive psychology, brain science, and human social history, combined with techniques such as cross-domain knowledge graph, causality inference, and continuous learning to establish effective mechanisms for stable acquisition and expression of knowledge. These make machines to understand and utilize knowledge, achieving key breakthroughs from perceptual intelligence to cognitive intelligence.
2. In-Memory-Computing addresses the “memory wall” challenges in AI computing
In Von Neumann architecture, memory and processor are separate and the computation requires data to be moved back and forth. With the rapid development of data-driven AI algorithms in recent years, it has come to a point where the hardware becomes the bottleneck in the explorations of more advanced algorithms. In Processing-in-Memory (PIM) architecture, in contrast to the Von Neumann architecture, memory and processor are fused together and computations are performed where data is stored with minimal data movement. As such, computation parallelism and power efficiency can be significantly improved. We believe the innovations on PIM architecture are the tickets to next-generation AI.
3. Industrial IoT will power the digital transformation
In 2020, 5G, the rapid development of IoT devices, cloud computing and edge computing will accelerate the fusion of information systems, communication systems, and industrial control systems. Through advanced Industrial IoT, manufacturing companies can achieve automation of machines, in-factory logistics, and production scheduling, as a way to realize C2B smart manufacturing. In addition, the interconnected industrial system can adjust and coordinate the production capability of both upstream and downstream vendors. Ultimately it will significantly increase the manufacturers’ productivity and profitability. For manufacturers with production goods that value hundreds of trillion RMB, If the productivity increases 5-10%, it means additional trillions of RMB.
4. Large-scale collaboration between machines becomes possible
Traditional single intelligence cannot meet the real-time perception and decision of large-scale intelligent devices. The development of collaborative sensing technology of the Internet of things and 5G communication technology will realize the collaboration among multiple agents — machines cooperate with each other and compete with each other to complete the target tasks. The group intelligence brought by the cooperation of multiple intelligent bodies will further amplify the value of the intelligent system: large-scale intelligent traffic light dispatching will realize dynamic and real-time adjustment, warehouse robots will cooperate to complete efficient cooperation of cargo sorting, driverless cars can cooperate to make the best tradeoff between efficiency and safety, and group UAV collaboration will efficiently get through the last kilometer of distribution.
5. Modular design makes chips easier and faster by stacking chiplets together
The traditional model of chip design cannot efficiently respond to the fast-evolving, fragmented and customized needs of chip production. The open-source SoC chip design based on RISC-V, high-level hardware description language and IP-based modular chip design methods have accelerated the rapid development of agile design methods and the ecosystem of open-source chips. In addition, the modular design method based on chiplets uses advanced packaging methods to package the chiplets with different functions together, which can quickly customize and deliver chips that meet specific requirements of different applications.
6. Large-Scale Production-Grade Blockchain Applications will Gain Mass Adoption
BaaS (Blockchain-as-a-Service) will further reduce the barriers of entry for enterprise blockchain applications. A variety of hardware chips embedded with core algorithms used in terminals, cloud and designed specifically for blockchain will also emerge, allowing assets in the physical world to be mapped to assets on the blockchain, further expanding the boundaries of the Internet of Value and realizing “multi-chain interconnection”. In the future, a large number of innovative blockchain application scenarios with multi-dimensional collaboration across different industries and ecosystems will emerge, and large-scale production-grade blockchain applications with more than 10 million DAI (Daily Active Items) will gain mass adoption.
7. A critical period before large-scale quantum computing
In 2019, the race in reaching “Quantum Supremacy” brought the focus back to quantum computing. The demonstration, using superconducting circuits, boosts the overall confidence in superconducting quantum computing for the realization of a large-scale quantum computer. In 2020, the field of quantum computing will receive increasing investment, which comes with increasing competition. The field is also expected to experience a speed-up in industrialization and the gradual formation of an eco-system. In the coming years, the next milestones will be the realization of fault-tolerant quantum computing and the demonstration of quantum advantages in real-world problems. Either is of a great challenge given the present knowledge. Quantum computing is entering a critical period.
8. New Materials Will Revolutionize the Semiconductor Devices
Under the pressure of both Moore’s Law and the explosive demand of computing power and storage, it is difficult for classic Si-based transistors to maintain sustainable development of the semiconductor industry. Until now, major semiconductor manufacturers still have no clear answer and option to chips beyond 3nm. New materials will make new logic, storage, and interconnection devices through new physical mechanisms, driving continuous innovation in the semiconductor industry. For example, topological insulators, two-dimensional superconducting materials, etc. that can achieve lossless transport of electron and spin can become the basis for new high-performance logic and interconnect devices; new magnetic materials and new resistive switching materials can realize high-performance magnetics Memory such as SOT-MRAM and resistive memory.
9. Growing Adoption of AI Technologies that Protect Data Privacy
The compliance costs demanded by the recent data protection laws and regulations related to the processing of personal data are getting increasingly higher than ever before. In light of this, there have been growing interests in using AI technologies to protect data privacy. The essence is to enable the data consumer to compute a function over input data from different data providers while keeping those data private. Such AI technologies promise to solve the problems of data silos and lack of trust in today’s data-sharing practices, and will truly unleash the value of data in the foreseeable future.
10. Cloud becomes the center of IT technology innovation
With the in-depth development of cloud computing technology, the cloud has grown far beyond the scope of IT infrastructure, and gradually evolved into the center of all IT technology innovations. The cloud has a tight relationship with almost all IT technologies, including new chips, new databases, self-driving adaptive networks, big data, AI, IoT, blockchain, quantum computing and so forth. Meanwhile, it creates new technologies, such as serverless computing, cloud-native software architecture, software-hardware integrated design, intelligent automated operation. In summary, cloud computing is redefining every aspect of IT. Cloud computing is continuously turning new IT technologies into accessible services and becoming the backbone of the entire digital economy.