Speaking at a keynote on the opening day of MWC Shanghai, David Wang (pictured), Executive Director of the Board, Chairman of the ICT Infrastructure Managing Board, Huawei, gave a speech entitled “Accelerating 5G-A Development and Opening the Mobile AI Era”.
June 2024 marks the fifth anniversary of 5G commercial use in China. In this time, the technology has had an unprecedented impact on the global mobile industry and achieved remarkable commercial results.
Firstly, 5G has expanded the breadth of global services. 320 commercial 5G networks have served more than 1.8 billion users, and the number of users is growing three times faster than 4G in the same period.
With 5G, average network speeds have improved tenfold, and services such as short video and livestreaming have been upgraded and refreshed. The first wave of 5G traffic accounted for more than 50% of the traffic in the market.
Finally, 5G has led digital transformation. The number of 5G private networks in the industry exceeds 30,000, driving the total economic output of nearly RMB20 trillion. In China, 5G is integrated into 74 of the 97 sectors, bringing economic and social benefits.
5G development is still accelerating. Just last week, 3GPP Release-18, the first version of 5G-A, was frozen.
At MWC earlier this year, more than 60 operators and partners confirmed that they would bring 5G-A into commercial use this year. 5G-A devices are ready – more than 30 mobile phones and CPEs will be fully supported this year. The changes brought about by 5G are taking place, and 5G-A will continue to drive the digital wave forward.
Looking back on the development of the mobile industry, the strides in development of networks and devices ushered in one great era after another. In 1991, 2G and mobile phones brought in the mobile voice era, enabling communications anytime and anywhere. In 2007, 3G and smartphones enabled the “mobile Internet era”, bringing global information within easy reach. This year will be the first for 5G-A commercial use and the first for AI on devices, marking the start of the mobile AI era which will see intelligent services become ubiquitous. We believe that the mobile AI era will bring about three important changes:
First, the transformation of human-computer interaction: the interaction mode will shift from “touch interaction” to “full-modal interaction,” supporting various interaction modes such as natural language, gestures, and emotion, providing a 24/7 experience.
Second, content production transformation: people will change from “retrieved content” to “AI-generated content.” Information production will change from prefabrication to customization, providing comprehensive and personalized real-time services.
Third, mobile terminal transformation: Mobile terminals will move from smartphones to “AI assistant terminals” and “embodied AI”, providing all-scenario intelligent entrances.
The great changes in the mobile AI era will inject new vitality into society and bring broad opportunities to the mobile industry. The first opportunity concerns every one of us. We can see that graphics interaction transformation has driven the rapid growth of mobile traffic over the past 10 years. The DoU increased from 140 MB in 2013 to 17 GB in 2023, a 120-fold increase in 10 years, doubling every 17 months on average. The traffic growth rate exceeds the Moore’s Law. In recent years, slowing traffic growth has become a common concern in the mobile industry. Can traffic growth return to “More than Moore”? We believe that the mobile AI era will lead to another transformation of content interaction and provide new impetus for traffic growth from the following three aspects.
First, interaction efficiency is improved. Generative will gradually replace retrieval as the mainstream. Take mobile shopping as an example: the time for consumers to select their favorite products is shortened from minutes to seconds, and the interaction efficiency is improved by a hundred times.
Second, the quality of interaction is improved. AIGC will promote the production of full-modal content such as 3D and spatial video. Videos of the same duration will be increased from 10+ MB to GB, and the amount of interactive information will be increased by 100 times.
Third, interactive object promotion. Each terminal has more than 100 built-in local small models, which collaborate with cloud-based foundation models in real time. The interaction objects of the terminal are expanded from a single human-machine interaction to hundreds of machine-machine interaction.
With the 100-fold improvement of interaction efficiency, quality, and objects, mobile traffic will grow explosively and return to “More than Moore” traffic growth.
The second key opportunity is in IoV. By 2030, more than 80 million new connected vehicles are expected to be added each year, and smart connected vehicles will subvert traditional driving methods and provide more pleasant and secure travel services.
Smart cockpits will improve people’s ride experience. The Gbps network experience supports the requirements of smart assistants, multi-screen viewing, and XR applications, making vehicles a “mobile third space” that integrates leisure, entertainment, and work. In the next two years, it is estimated that the duration of in-vehicle video viewing will increase by 10 times, and in-vehicle traffic will increase by 100 times.
Smart driving will subvert the traditional driving mode. Through 100 GB data upload training every month, the smart driving system continuously improves model accuracy and accurately identifies and responds to diversified traffic environments. In addition, millisecond-level inference based on vehicle-road collaboration provides stable and reliable driving policies in real time, increasing the traffic speed by more than 20% and reducing traffic accidents. Smart connected vehicles will realize the value of IoV experience, further bring time value and security value, and reshape the value of connections.
Third, the mobile AI era will greatly expand the boundaries of the Internet of Everything. At present, mobile networks are accelerating the development of the Internet of People and Things, serving individuals, families, and all industries. By 2030, mobile networks will carry 10 billion people. As machines gain human-like thinking and mobility capabilities, network connections will expand from “carbon-based human” to “silicon-based human”, creating 10 billion new connections.
For individuals, everyone will have an AI assistant in the future, running across mobile phones, glasses and new AI terminals. AI digital human will evolve from tools to partners, responding to requirements in life, work, and learning scenarios in real time. It is estimated that the number of AI assistant terminals will reach 5 billion by 2030. Compared with human users, each AI assistant generates more than 10 times the traffic. Through device-cloud collaboration, it supports learning-based evolution, and intelligent interaction services.
For the industry, each smart factory will be equipped with an AI brain to drive AI machine employees into all production links. AI machine employees will be capable of self-learning and accurate response, enabling production, maintenance, quality inspection, and logistics to improve quality and efficiency.
“Digital humans” and “robots” will break the traditional connection ceiling, create tens of billions of “new connections of people” and lead a new era of human-machine fusion.
Faced with the broad opportunities of the mobile AI era, the industry needs to work closely to cultivate fertile soil for innovation and grasp the development dividends of the times.
As an important pillar of the mobile AI era, 5G-A needs to accelerate its development and work with operators and industry partners to create new business value.
Accelerating 5G-A development from the dimensions of “Networks for AI” and “AI for Networks”
Networks for AI is a solid foundation for the mobile AI era; diversified services pose higher requirements on network capabilities such as large downlink, large uplink, and low latency. Take bandwidth as an example: to achieve immersive experience in the 3D digital intelligence world, experience devices must support binocular 8K resolution, 120 fps high frame rate, and provide six degrees of freedom for free interaction. Such experience requires a guaranteed downlink rate of at least 1 Gbps per person.
On the latency side, achieving real-time interaction with AI assistants requires end-to-end latency of within 400 ms. In multi-modal interaction scenarios involving voice and video, more than 2000 tokens are generated at a time. Considering that the large model requires at least 320 ms to process the data, and the processing time of the transport network and terminals, the network air interface delay must be less than 20 ms.
Meeting the first upgrade of people-vehicle connections will require accelerating the evolution of all frequency bands to 5G-A, fully unleashing the value of each frequency band, and building the foundation for ultimate network experience.
On the basis of full-band pooling, technologies such as multi-carrier aggregation and MBSC are used to aggregate spectrum resources on demand, and sufficient 5G-A pipe capabilities are built with ultra-large bandwidth, meeting everyone’s requirements for gigabit and ultimately 10Gbps.
It’s also crucial to make full use of the advantages of different frequency bands and use technologies such as uplink-downlink decoupling and FSA to implement multi-frequency mutual assistance, achieve the ultimate multi-dimensional network experience, and meet the requirements of uplink gigabit transmission and ultra-low latency less than 20 ms.
In addition, AI for Networks enables network productivity improvement with intelligence. AI capabilities must be injected into the NE, network, and service layers of wireless networks to achieve intelligent performance, simplified O&M, and intelligent service operations.
With the evolution of the network and the development of services, the complexity of the network increases significantly. It is imperative to fully embrace AI and improve network O&M efficiency.
The big communication model accelerates the development of network automation towards AN L4. It provides two types of applications: role-oriented copilots, which improves personnel efficiency through professional knowledge Q&A and O&M assistance, and scenario-oriented agents.
The two types of applications will completely reshape the operation, optimization, and maintenance of telecom networks.
Service operation scenario: Foundation models simulate multi-modal data such as topology, environment, and traffic distribution in real time, helping carriers quickly and accurately provision numbers.
Network optimization scenario: Foundation models control multi-dimensional indicators such as the network rate, coverage, and energy efficiency in real time, and dynamically formulate a multi-objective comprehensive optimization policy through the “wireless agent” to optimize experience and power consumption.
Troubleshooting scenario: By understanding machine data, agents automatically plan and orchestrate tasks, significantly reducing the number of alarms by 90% and resolving software commissioning problems by 40%. For the remaining 60% of hardware problems, copilots assist engineers to perform accurate site visits and quickly solve problems.
In the mobile AI era, a solid network foundation provides fertile ground for innovative applications and business models. These will help to improve the value of all-scenario connections, upgrade business models, and create new business value.
For the ToC scenario, the continuous innovation of intelligent applications and the improvement of consumer requirements for multi-dimensional experience drive carriers to transform from single-factor traffic monetization to multi-factor experience monetization.
The next decade will be a decade to embrace the mobile AI era, a decade to accelerate the intelligent world, and a decade for mobile networks to play a cornerstone role.
Huawei looks forward to working with the industry to jointly seize the huge opportunities in the mobile AI era and build a better intelligent world.
MORE ARTICLES YOU MAY BE INTERESTED IN…
This post was originally published on the 3rd party mentioned in the title ofthis site
