On April 12th, 2021, at the Huawei Global Analyst Conference, Huawei’s Director and Director of the Strategic Research Institute Xu Wenwei announced 9 major technical challenges and research directions towards the Smart World 2030.
The following is the full text of Xu Wenwei’s speech: (Translated)
Hello, ladies and gentlemen, and welcome to the 18th Analyst Conference.
In the past year, the epidemic, globalization, and the entire world have experienced tremendous challenges. Today, we are standing at the starting point of the next decade, with unknowns and longings. The ICT industry is also facing new challenges and urgently needs a new round of breakthroughs.
Population and energy are the two major themes of the development of human society:
The United Nations report shows that by 2030, there will be 8.6 billion people in the world, with more than 12% of people over 65 years old, and the proportion of people under 25 years old will continue to decline. The aging population and insufficient labor force have become challenges for social development. People’s pursuit of health, hope to live well, live long, and walk safely.
In addition, global energy consumption is increasing at an annual rate of 1.7%. The report shows that since the 18th century, human energy consumption has increased by 22 times, of which fossil energy accounts for as much as 85%. Sustainable development of energy is a problem facing us.
We predict that in 2030, renewable energy will account for more than 50%; electric travel will become the main force, and electric vehicle sales will account for more than 50%; AI will change everything, and the utilization rate of home intelligent robots will exceed 18%. ICT technology has the potential to help reduce 20% of global carbon emissions by empowering other industries in the next ten years.
We hope to get rid of physical limitations and improve perception. Although mobile phones have now reached 100x zoom, there is a huge gap between them and the biological world. For example, spiders far surpass the human eye in terms of object contours and motion calculations. So, can you learn the eyes of spiders? We can create better cameras that meet the needs of autonomous driving.
We hope to surpass the wisdom of biology and develop new types of computing. Nowadays, artificial intelligence is widely used, but deep neural network training is difficult and consumes a lot of power. Sometimes it can’t compare with ants. Ants can do a lot of things with 0.2 milliwatts of power consumption, including building nests, making friends, even fighting and raising aphids, etc. Can we learn and learn from the working methods of living things in depth, and start development from the realization of simple intelligence?
We hope to overcome the obstacles of space and realize the immersive experience. The current 5G communication is far from satisfying the demands of immersive communication. We need to develop a faster and lower-latency network to support human-level holographic communication.
We hope to expand the limits of cognition and develop mesoscopic devices. Scientists use computational methods to achieve molecular and atomic design and assembly. In this way, the performance of chips and devices can be greatly improved.
Matter, energy, and information are the three elements of the world, and they are the starting point for us to grasp the future challenges and directions. Matter is the existence of origin, energy is the existence of movement, and information is the existence of connection.
In the next ten years, the number of connections will reach hundreds of billions, the broadband speed will reach 10Gbps per person, the computing power will be increased by 100 times, the storage capacity will be increased by 100 times, and the use of renewable energy will exceed 50%. Around the generation, transmission, processing and use of information and energy, technology needs to evolve continuously.
Based on these predictions and assumptions, next, I will talk about the challenges and development directions in the next ten years.
Challenge 1: Define 5.5G to support the future diversified connection of 100 billion scale
The first challenge is the challenge of the Internet of Everything. We must not only connect all people, but also connect a large number of things, and the needs of connecting things are diverse.
Currently, the three major scenarios defined by 5G are difficult to support diverse IoT scenarios. For example, the application of the Industrial Internet of Things requires both massive connections and large uplink bandwidth. A scene must be added between eMBB and mMTC, named UCBC (uplink ultra-wideband); there is a category of applications that require both ultra-wideband and low bandwidth. For latency and high reliability, a scenario must be added between eMBB and URLLC, named RTBC (Real-time Broadband Interaction); vehicle-road collaboration in the Internet of Vehicles requires both communication capabilities and perception capabilities, and HCS scenarios must be added (Communication perception fusion).
Therefore, it is necessary to change from the “triangle” of the 5G scene to the “hexagon” of the 5.5G scene, from supporting the interconnection of all things to enabling the intelligent connection of all things.
Challenge 2: Controlling light at the nanometer scale and achieving exponential growth in optical fiber capacity
The challenge of 5G connection is in quantity, and the challenge of fiber connection is in capacity. Today, a single fiber carries 1 million people to watch 4K video, and in 2030 it will carry 1 million people to enjoy MR (mixed reality), and the single fiber capacity will increase 10 times, exceeding 100T.
The first is the optical transceiver laser, which uses high modulation devices to achieve a 2 to 3 times increase in baud rate; at the same time, new modulation codes and algorithms are used to double the capacity. Thin-film high-bandwidth modulators are the development direction.
Secondly, we must develop broadband, low-noise, artificially controllable new optical amplifiers to achieve reliable ultra-long-distance transmission; the key technology is optical amplifiers close to the quantum limit.
The third is the dynamic control capability of the optical network. The wavelength division network is transformed into a “synchronous” system, which improves the anti-interference ability and realizes the efficient use of optical resources through calculation. The microcavity optical frequency comb is the key.
In the more distant future, it is also necessary to study new types of optical fibers and optical systems such as SDM (Space Division Multiplexing) to achieve a hundredfold increase in single-fiber capacity.
Challenge 3: Towards industrial interconnection, network protocols must be optimized
Today, the main body of network support is tens of billions of consumer interconnections. In 2030, the main body of network support is trillion-level industrial interconnection, and network protocols face three tests.
The first is certainty. The deterministic delay guarantee capability is needed, and the current best-effort network delay can be converted into a definite delay that can be calculated in advance through the “new network calculation theory and agreement”.
The second is safety. In the context of the Internet of Everything, the security defense system poses serious challenges. A large number of external devices such as drones, cameras, edge computing, and sensors have introduced new insecure factors, and an end-to-end endogenous security framework and protocol must be built.
The third is flexibility. The needs of thousands of industries are diverse. Some need a longer IP address, and some need a shorter IP address. The fixed-length IP address must be extended to a new IP protocol that can flexibly define semantics and syntax.
Challenge 4: General computing power is far from keeping up with the needs of the smart world, and supercomputing power must be built
In the smart world, connectivity determines the breadth, so calculation determines the strength. Facing 2030, the demand for computing power will increase by 100 times. But at present, the annual improvement rate of single-core CPU performance has dropped from 50% to 10%, and general-purpose computing is inefficient in certain areas. How to create super computing power is a huge challenge.
First, digital computing is moving from general purpose to special purpose, to heterogeneous computing where multiple computing architectures coexist, and various CPUs, GPUs, and XPUs coexist.
Second, analog computing will show advantages in specific areas. Photonic computing will be applied to signal processing, combinatorial optimization, machine learning and other fields, especially for wireless Massive MIMO and optical communications. There will be great application scenarios.
Challenge 5: Efficiently extract knowledge from massive multi-modal data to achieve a key breakthrough in industry AI
The intelligent world is inseparable from AI, and the issue of AI application fragmentation and AI’s credibility cannot be avoided.
The versatility of the AI model is the key to solving application fragmentation. Through a large amount of unlabeled data and a larger model, from full supervision to self-supervision, building a general AI system is a direction that needs to be broken.
Secondly, the convergence of AI and scientific computing also provides a great use for AI applications to get out of fragments. AI brings new ideas, new methods, and new tools to scientific computing, and the rigorous system of scientific computing also helps improve the interpretability of AI.
Trustworthy AI is our long-term goal. Especially in key areas where human life is critical, such as unmanned driving, the problems from relevance to causality must be solved.
Challenge 6: Break through the von Neumann limit and build a new type of storage with a hundredfold increase in density
Storage faces two major problems: it can be saved and used well
First, it must survive. The storage density per unit space and energy consumption has to be increased by 100 times, but the current media technology is limited by technology and power consumption, which cannot be supported. In the future, storage systems must break through new large-capacity and low-latency memory technologies, breakthrough ultra-large-capacity media technologies such as DNA storage and high-dimensional new optical storage, breakthrough ultra-large storage space models and coding technologies, and break the capacity wall.
Second, use it well. In the future, the data access bandwidth of the storage system will be from TB level to PB level, the access delay will be reduced from ms level to us level, and the performance density must be increased by a hundredfold. Under the Von Neumann architecture, data must be moved between CPU, memory, and media, but the current bandwidth speed of PCIE and DDR is far from keeping up with the performance growth of external networks. In the future, storage systems will break through the limitations of the von Neumann architecture, shift from CPU-centric to memory-centric, data-centric, and from moving data to moving computing, breaking the performance wall.
Challenge 7: Combine computing and perception to achieve a multi-mode interactive surreal experience
The smart world needs to create the ultimate user experience. I think that in 2030, surreal experiences will become a reality.
Surreal experience, which requires a seamless integration of the virtual world and the real world. And can accurately perceive and restore the world, and understand the user’s intentions in a world that combines virtual and real. The senses of hearing, sight, touch, and smell must be opened up to realize multi-mode interaction between people and hundreds of edge devices.
In order to achieve this goal, it is necessary to treat the user’s environment as a supercomputer, relying on multi-mode sensors such as language, touch, light perception, and brain-computer to collect and transmit information, recognize the user’s intention, and use naked-eye 3D, holographic projection , AR contact lenses, digital smell and digital touch technology are presented to users.
Challenge 8: Realize active health management through continuous health monitoring
The aging population has brought more chronic diseases. According to statistics, 85% of deaths are due to chronic diseases, and chronic diseases must be detected in real time. It is necessary to overcome the need for medical-grade wearable devices, such as non-invasive blood sugar, continuous blood pressure, continuous ECG and other vehicles.
Taking blood pressure detection as an example, optical sensors can provide more accurate pulse waves than PPG, and provide higher quality data input for blood pressure modeling and algorithms. Combining cloud services and artificial intelligence technology to create a complete personal health big data platform for individuals to achieve active health management. Through brain-computer interfaces, electromyographic interfaces, wearable robots, etc., from being taken care of to autonomous management, the happiness of the elderly is improved.
Challenge 9: Build a smart energy Internet to realize green power generation, green power storage, and green power consumption
The current “carbon peak and carbon neutrality” is accelerating the transition to new energy, and it also brings new challenges in power generation, energy storage and electricity use.
From the perspective of power generation, the evolution from centralized to distributed means that the power generation system is closer to users. In the past, it was a pure electricity use scenario. In the future, it will also have the ability to generate self-generation. Network characteristics; the volatility, multi-energy complementary characteristics of new energy power generation, and intermittent power supply characteristics make new energy the main power source, which poses huge challenges.
From the perspective of energy storage, in the past, there were only power generation and electricity consumption, and how much energy was used. In the future, new energy-based power generation must have a buffer pool for energy storage, which makes the network more complicated. It is necessary to achieve low-cost, zero-carbon large-scale energy storage, and to maximize the use of green electricity through intelligent dispatch.
From the perspective of electricity consumption, comprehensive smart energy must be promoted to realize residential/building/factory energy management systems, zero-carbon communities, zero-carbon parks, and zero-carbon cities.
Therefore, it is necessary to build a smart energy Internet to realize green power generation, green energy storage and green electricity, which involves several key technologies:
First, management technology. Big data, AI, cloud and other ICT technologies are integrated with the energy Internet, through the energy cloud + energy network, to achieve bit management watts.
Second, control technology. Through the power, electronic energy router, the two-way energy flow and the intelligent distribution of power are realized, and the intelligent controller of the energy network is constructed.
Third, energy storage technology. Develop new energy storage technologies, such as new electrochemistry, hydrogen energy, etc., to meet the energy storage needs of different scenarios.
Fourth, the basic technology of power electronics. A new type of compound power semiconductors, including SiC/diamond for medium and high voltage and GaN technology for medium and low voltage, realize further efficiency and miniaturization of energy components.
The above are the nine major technical challenges and research directions we put forward from the perspective of the ICT industry. They are also our expectations for the Smart World 2030. We hope to achieve stronger connections, faster computing, and greener energy.
Overcome challenges with an open, inclusive and collaborative innovation mechanism
In order to meet the needs of human development and solve the problems we face, we need to pool the wisdom and innovation capabilities of all mankind, and we must overcome challenges with an open, inclusive and collaborative innovation mechanism. The industry must work closely with universities and scientific research institutions, and use industry challenges and world-class problems to guide the direction of scientific research.