Striding Towards the Intelligent World 2030
—Technological Challenges and Research Directions
At Huawei Global Analyst Summit 2021, Director of Board and President of Huawei's Institute of Strategic Research, William Xu, shared the company's vision for an Intelligent World in 2030, highlighting the nine technological challenges and research directions that will help bring it to fruition.
At Huawei Global Analyst Summit 2021, Director of Board and President of Huawei's Institute of Strategic Research, William Xu, shared the company's vision for an Intelligent World in 2030, highlighting the nine technological challenges and research directions that will help bring it to fruition. Key to this vision is greater collaboration between industries, academia, and research institutions, alongside an open, inclusive, and collaborative approach to innovation that unleashes the full potential of human creativity to address the challenges we all face.
William Xu, Director of Board and President of Huawei's Institute of Strategic Research
The following is the full transcript of his speech:
Ladies and gentleman, welcome to the 18th Huawei Global Analyst Summit. We have all had a difficult year, facing many new challenges because of the pandemic and anti-globalization. Today, we are looking at the beginning of a new decade that will bring many new uncertainties and opportunities. The ICT industry will also face new challenges and need to make new breakthroughs.
Population and energy underpin the well-being of our society
A UN report shows that by 2030, the global population will reach 8.6 billion, of which over 12% will be 65 or older. The percentage of people aged 25 years or younger is also decreasing. Ageing populations and labor shortages are hindering social progress. There is an increasing need for solutions that not only prolong life, but also improve quality-of-life and ensure dignity during end-of-life care.
Global energy consumption is simultaneously growing at an annual rate of 1.7%. Energy consumption has increased by 22 times since the 18th century. Currently 85% of the energy comes from fossil fuels. Energy sustainability is a daunting challenge that faces all of us.
The path to sustainability: Low-carbon, electric, and intelligent
Low-carbon energy, broader electrification of industries, and intelligence are the path to sustainability.
By 2030, more than 50% of all energy will come from renewable sources, more than 50% of cars sold will be electric, and more than 18% of homes will have smart robots.
By empowering a wide range of industries, ICT technology has the potential to reduce global carbon emissions by 20% over the next decade.
We have many expectations for the future that will require us to transcend our limits.
We hope we can exceed the biological limits of perception. Now phone cameras have achieved 100x zoom, but there is still a huge gap between camera capabilities and what we see animals achieve in the natural world. For example, spiders are much better at detecting the outlines of objects and motion, and we may learn from them to develop cameras that meet the requirements of autonomous driving.
We hope we can go beyond biological intelligence as we develop new computing technologies. Though we now have many AI applications, deep neural networks are hard to train and consume large amounts of energy. They cannot fulfill functions as efficiently as the brains of ants. An ant's brain only consumes 0.2 mW of power, but can process many activities, like nesting, making friends, fighting, and feeding aphids. I think we can learn from such creatures to develop AI by starting with building intelligence in simple scenarios.
We hope to transcend physical limits to enjoy truly immersive experiences. Existing 5G networks are far from being able to offer such experiences. So we need to offer faster network speeds and lower latency for life-size holographic communication.
We also hope we can extend our horizons by developing mesoscopic devices. Scientists use computing to realize molecule- and atom-level design and assembly to greatly improve the performance of chips and components.
Our world is constructed based on three pillars – matter, energy, and information. Together, they determine how the world works. And from them, we can figure out what future challenges we will face. Matter is the origin of existence; energy drives motion; and information determines connections.
By 2030, there will be hundreds of billions of connections around the world. Broadband speeds of 10 Gbps will be available to every user. We will see a 100-fold increase in computing power and storage capacity. More than 50% of energy will come from renewable sources. The technologies that power the generation, transmission, processing, and use of information and energy will need to evolve.
Based on these predictions and assumptions, I want to talk about the challenges we will need to address and directions of development in the next decade.
Challenge 1: Defining 5.5G to support hundreds of billions of diverse connections
The first challenge will be in connecting all things. In addition to connecting people, we need to also connect massive numbers of things. The demands for those connections will be very diverse.
The three use cases defined by 5G are not able to support some of the more diverse IoT scenarios. For example, industrial IoT applications require both massive numbers of connections and large uplink bandwidth. So they need Uplink Centric Broadband Communication (UCBC) – a scenario that falls between eMBB and mMTC. There is another type of applications that need ultra-broadband, low latency, and high reliability. They require Real-Time Broadband Communication (RTBC). This is a scenario that falls between eMBB and URLLC. Vehicle-road collaboration in connected vehicles requires both communication and sensing capabilities. So we need another new scenario, Harmonized Communication and Sensing (HCS).
5.5G must cover these three new scenarios currently not covered by 5G: UCBC, RTBC, and HCS. Together they will take us beyond the connection of everything, enabling an intelligent connection of everything.
Challenge 2: Nanoscale optics for an exponential increase in fiber capacity
The challenge that hinders 5G connectivity will lie in the quantity of our connections, while the challenge that hinders fiber connectivity will lie in fiber capacity.
Today, a single fiber can simultaneously support 1 million 4K video streams. In 2030, it will need to be able to support 1 million people interacting in mixed reality. This means that, for the capacity of a single fiber to exceed 100 Tbps, it will need to expand by 10-fold.
First of all, we will have to work on optical transceiver lasers and use high-modulation components to double or triple baud rates. In addition, new modulation coding and algorithms will be required to multiply the capacity. Thin-film high-bandwidth modulators will be the way forward.
Second, we must develop new, broad-frequency, and low-noise optical amplifiers that support manual control for reliable, ultra-long haul transmissions. The key technology will be optical amplification that brings us close to the quantum limit.
Third, we will need to study dynamic controls for optical networks and transform the WDM network into a synchronous system to improve anti-interference features and efficiently use optical resources through computing. The key technology will be micro-cavity optical frequency combs.
Longer term, we will also need to research new fiber and optical systems like Space Division Multiplexing (SDM) to increase the capacity of a single fiber by 100-fold.
Challenge 3: Optimizing network protocols to connect all things
Today, our primary networks can support tens of billions of consumer connections. By 2030, they will need to support trillions of industry connections. This will bring three major obstacles to network protocols.
The first will be in achieving deterministic networks. These networks need guaranteed deterministic latency. We must use new network calculus theories and protocols to transform the best-effort network latency we have today into a deterministic latency that can be calculated in advance.
The second obstacle will be in security. When all things are connected, security systems will face serious challenges. Large numbers of devices like drones, cameras, edge computing devices, and sensors will all present new security risks. The time is ripe to develop intrinsic, end-to-end security frameworks and protocols.
The third obstacle we will face is in flexibility. As the variety of industry requirements increases, some will require longer IP addresses, while others will require shorter. To resolve this issue, we will need to expand IP addresses with fixed lengths and develop new Internet protocols that feature sematic and syntax flexibility.
Challenge 4: Advanced computing power strong enough for the intelligent world
If connectivity determines the breadth of the intelligent world, computing will determine its depth.
In 2030, the demand for computing power will see a 100-fold jump. In the past we would see single-core CPU performance increase about 50% every year. But now, that rate has dropped to 10%. We are also finding that general-purpose computing is very inefficient in some domains. Providing sufficiently advanced computing power will be a huge challenge.
First, we must move digital computing from general-purpose computing to special-purpose computing and then to heterogeneous computing that allows for the coexistence of multiple computing architectures like CPUs, GPUs, and xPUs.
Second, we will need to leverage the benefits of analog computing in special-purpose domains. Photonic computing will be used in domains like signal processing, combinatorial optimization, and machine learning. In particular, photonic computing has huge application potential in Massive MIMO and optical communication.
Challenge 5: Extracting knowledge from massive data for breakthroughs in industrial AI
The intelligent world would not be possible without AI, so the next challenge is about the fragmentation of AI applications and AI trustworthiness.
We believe the key to addressing the issue of fragmentation will be to develop general-purpose AI models. General-purpose AI systems can be achieved by using large amounts of unlabeled data and larger models, and shifting from supervised to self-supervised learning. Making breakthroughs in these areas will be critical.
In addition, we should bring AI and scientific computing together. This will also help address the fragmentation of AI applications. AI will bring new approaches, methods, and tools for scientific computing, while a rigorous scientific computing system will help make AI much more explainable.
AI trustworthiness is our long-term goal. It is particularly important in key domains that relate to matters of life and death, such as autonomous driving, where we must address daunting challenges from relevance to causality.
Challenge 6: Going beyond von Neumann architecture for 100x denser storage systems
The sixth challenge I want to talk about is storage.
Storage capacity and performance will be two issues that need to be addressed for future storage systems.
First, we need much higher storage capacity. The capacity density will need to be 100 times higher than what we currently have. Existing storage media cannot achieve this level of capacity density though due to process and power consumption restrictions. To overcome this capacity hurdle, we will need breakthroughs in new technologies, including large-capacity and low-latency in-memory computing technologies, ultra-large capacity media technologies such as DNA storage and high-dimensional optical storage, and ultra-large storage space model and coding technologies.
Second, we will also need significantly improved storage performance. As the data access bandwidth of storage systems increases from TBs to PBs, and access latency drops from milliseconds to microseconds, we will need performance density to increase 100 times beyond what we have today. Under the von Neumann architecture, data needs to be transmitted between CPUs, memory, and storage media. The PCIe and DDR bandwidths we currently have will not be able to keep up with network performance growth. To breakthrough this performance wall, we will need to move past the von Neumann architecture and shift away from CPU-centric storage and towards memory- and data-centric storage. We will also need to focus more on computing migration rather than data migration.
Challenge 7: Combining computing and sensing for a hyper-reality, multi-modal experience
Challenge number seven will be creating an inspired user experience that will be an integral part of the intelligent world. This kind of hyper-reality will become true reality by 2030.
Hyper-reality experiences can be achieved when the virtual world is seamlessly integrated with the physical world, and when the virtual world can accurately perceive and render the physical world while understanding user intent in mixed reality.
Hearing, vision, touch, and smell must all be integrated to enable multi-modal interactions between individuals and hundreds of edge devices.
To achieve this goal, the integrated user environment needs to work like a super computer. Multi-modal sensors are needed to collect and transmit language, touch, light perception, neural, and other types of information, as well as to perceive user intent. Technologies like naked-eye 3D, holographic projection, AR contact lenses, digital smell, and digital touch will be needed to display this information to users.
Challenge 8: Enabling continuous self-monitoring for more proactive health management
Ageing populations will lead to uptick in chronic illness.
85% of deaths are currently attributed to chronic illness. Effectively treating chronic illness requires real-time monitoring. This in turn will require medical-grade wearables to achieve non-invasive blood glucose monitoring, and continuous blood pressure and ECG monitoring. For example, optical sensors can provide more accurate pulse waves than PPG sensors. They can also offer higher quality data for blood pressure modeling and algorithms. We should look into building a complete personal health big data platform based on cloud services and AI technologies to enable proactive health management. With the support of brain-computer interfaces, sEMG interfaces, and wearable robots, we can empower the elderly with the capacity to proactively manage their own health.
Challenge 9: An intelligent Internet of Energy for the generation, storage, and consumption of greener electricity
New carbon neutrality and emissions peak targets have accelerated a global transformation towards new energy. But this transformation also brings new challenges to the fields of electricity generation, energy storage, and electricity consumption.
Electricity generation systems have moved closer and closer to individual users as centralized generation has evolved into distributed generation. In the past, only consumption happened on a user-level. In the future, electricity generation will also happen on-site. Before this can happen though, there will need to be more bidirectional energy codes and the electrical grid will need to be more like an Internet. Electricity generated from new energy sources is a more volatile intermittent resource, due to the complimentary nature of multi-energy systems. Huge challenges must be resolved before new energy becomes a primary energy source.
While previously we primarily only had to worry about generation and consumption, new energy makes energy storage buffer systems just as critical. We will no longer just instantaneously generate the amount of electricity needed to simply meet user demand. This will make our energy grids much more complex. We must figure out how to store large amounts of energy at low costs and with zero carbon emissions, and how to maximize the use of green electricity through intelligent scheduling.
For energy consumption, we must promote integrated smart energy to build energy management systems for households, buildings, and factories, and to create zero-carbon communities, campuses, and cities.
An intelligent Internet of Energy must be built to achieve green electricity generation, storage, and consumption. This will require advancements in several key technologies.
The first will be in management technologies. ICT technologies such as big data, AI, and cloud need to be integrated with the Internet of Energy to achieve bit-based watt management through an energy cloud and energy network.
The second will be in control technologies. Power electronics-based energy routers can be used to build intelligent energy network controllers that realize bidirectional energy flow and intelligent energy distribution.
The third will be in energy storage technologies. New energy storage technologies, including new electrochemical and hydrogen storage mediums, need to be developed for multiple scenarios to meet these growing storage requirements.
The fourth will be in power electronic technologies. Wide-bandgap semiconductors, including SiC and diamond for medium- and high-voltage applications, and GaN for medium- and low-voltage applications will be needed to make energy components more efficient and compact.
These are the nine technological challenges and directions for further research that we believe are crucial based on our experience in the ICT industry. They also represent what we believe is needed to achieve an intelligent world by 2030: Stronger connectivity, faster computing, and greener energy.
Overcoming challenges through an open, inclusive, and collaborative approach to innovation
We need to pool the wisdom and innovation capabilities of all people to continue driving human development and tackle the massive challenges facing us all. We must overcome challenges by taking an open, inclusive, and collaborative approach to innovation. Industry players must collaborate closely with universities and research institutions, and inform and guide scientific research by defining universal and industrial challenges.
People have always imagined what the future will be like. But with technology we can actually get there. We need to integrate industrial challenges and academic insight, then adopt a venture capital mindset to innovate together and build the Intelligent World of 2030. Thank you.