5G-A, AI, and robots: the tech ecosystem is racing ahead

Interview 1: In China, It's “AI with Everything”

Alex Sinclair, CTO of GSMA

About six years into the 5G cycle, China has rolled out 5G on a broad scale and is leading the way in development of AI as well. What has enabled China’s mobile industry to be so far ahead in 5G and AI?

The Chinese were one of the first out of the gates for 5G standalone and, of course, 5G-Advanced. That certainly helps the scale of the market. Nearly 95% coverage of the population, 300 5G base stations for every hundred thousand people—that's pretty amazing.

And we shouldn't forget the enabling environment the government has created, making things a lot easier to deploy sustainably, and the willingness of people, businesses and society as a whole to adopt the technology. It’s a powerful mix.

What opportunities and benefits does 5G-Advanced enable?

We bundle them into three categories: performance enhancements, efficiency enhancements, and new use cases.

One example is China Mobile, Huawei, and Leju Robotics (see interview below with Leju CEO Chang Lin). Leju has developed a humanoid robot that performs fine-grained positioning.

And then, of course, if you're looking at the 5G-Advanced network itself, I was fortunate enough to be in Beijing last year where they launched the largest 5G-A deployment to 10 million people, with downlink speeds of 11.2 Gbps and uplink speeds of 4 Gbps. That's pretty amazing. Imagine what you can do with that.

What are some of the more transformative impacts of AI?

Well, we're seeing AI with everything at the moment. But China Mobile has been combining with Huawei to use IoT sensors and AI for quality control on manufacturing lines. China Telecom has a completely different application for rural healthcare, focusing on medical imagery analysis. And China Unicom is predicting faults for steel plants in Hebei Province.

What needs to happen in order for 5G-A and AI to realize their true potential?

First, don't take your eyes off the prize, because here in China alone, by 2030, we're predicting US$2 trillion of value to the economy.

Basically, you need investment, spectrum, and partnership. As for investment, in China alone we're expecting the data to triple by 2030. And we'll see similar developments in other countries. So basically, keep investing.

In terms of spectrum: to keep up with the network traffic, you're going to have more low-, mid-, and high-band spectrum, including the all-important 6 GHz band.

But if there's one thing our industry knows more about, it's partnerships. And we're starting to see the right kind of partnerships in new areas like AI as well. We just launched a mobile AI community in Barcelona, for example. We're doing open telco LLM benchmarks together, and we've got a “responsible AI maturity model.”

And one of my personal favorites: we've just started a new project on agent interoperability as well. So, there's a lot going on.

Interview 2: Embracing the Robot Era”

Chang Lin, CEO of Leju Robotics

How can robots transition from the training in simulated environments to real-world industrial applications?

Chang Lin: We use AI methods to train the robots. So, we want to collect data from the robots and put it into large models. We can then refine our methods of controlling and manipulating the robot for commercial use.

As robots increasingly integrate into production lines and even daily life, what are some of the industry synergies we need to see? For example, does a lot of work need to be done on standards?

Chang Lin: Yes. Technical standards are at a very early stage, so it’s difficult to clarify the standards right now. Another major challenge is the network. For example, when we put one robot into a factory, it works very well with Wi-Fi. But if we use five robots, there are many errors because the network cannot handle it.

So, we collaborated with Huawei on 5G-A solutions for the robots. When we put 5G-A onto the robot, five robots work very well—or even 30 or 50 robots.

Do people need to get used to having robots around all the time?

Chang Lin: Yeah. Some people say it is frightening to meet the robots. But it is necessary for us to embrace the robot era. Maybe we can work or live with robots because they can be a partner—or even a friend.

Leju Robotics has had a busy year already. What are some of the milestones so far?

Chang Lin: At MWC-Barcelona in March, together with China Mobile and Huawei, we launched the world’s first humanoid robots equipped with 5G-A connectivity. Our robots are being used in the factories of First Auto Works (FAW), a major Chinese car manufacturer. So, we use the 5G-A network to make the robots more intelligent.

Also, there is a famous race in China using robots. Some older robots use the remote controller because the network is not available for the robots to control themselves. When we have 5G-A—maybe in the next year—we can control the robots without remote control. I think this is huge progress, and very exciting.

Interview 3: Partnership Paradigm

Shaun Collins, Executive Chairman and Founder, CCS Insight

Do you think the operators should build or partner for their own AI capabilities?

I think operators, even those more advanced in developing AI within their portfolios, have all chosen to partner at least on some level.

That's emblematic of what 5G is, because, if you look at 3G and 4G, we didn't really need big partnerships to make things happen. They either happened independently, or as a result of 4G just appearing. But what's become clear with 5G across the world is that, in order to extract the maximum value of it, it is basically a series of partnerships that will do it.

At CCS Insight, we've been saying for a while that 5G is a team sport: if you're going to succeed, you have to do it with others. That's easy to say and hard to do if you're trying to put together allegiances and groups of enormous companies that are used to having their own way.

But I think [the question] whether you're going to do AI on your own, or as a partnership— for an operator, that decision was made three years ago.

Can AI generate new revenue streams in the future?

I'd be surprised if it didn't. In the enterprise market, 5G and AI are a very potent combination. When you add in Cloud and security, and some of the IoT elements we're seeing, it becomes a really potent group of opportunities for operators and their partners.

Consumer is slightly more challenging. Because 4G was so good, 5G hasn't been enough of an improvement for most vendors or users around the world to go, “Oh, yes, I want 5G please, and I'll pay more for it.”

What we're having to do is create experiences that allow them to identify the worth of that extra connectivity or that low latency. So, what you're seeing is operators building new experiences, for which they'll try and charge a little bit more. So, it's not quite the same as it was in 4G, but it's still there.

Interview 4: Beyond scale: how AI will become smarter, cheaper, and more personal

Liu Zhiyuan, Associate Professor of Tsinghua University, Co-founder and Chief Scientist of ModelBest

For the future of AI, how should we assess the trade-off between parameter scale and efficiency?

Liu Zhiyuan: Over the past six years, we have witnessed the continuous expansion of the parameter scale of large models, which was initiated by OpenAI's scaling laws. Models cannot keep growing indefinitely, because both the training and usage costs are directly proportional to model parameter scales.

Entering the era of intelligence means making large models accessible and usable for all. Technological innovations are needed to enable AI to provide more with smaller parameter scales. This is the main reason for our commitment to developing high-density large models. The future of large models will not simply be about making parameter scales larger, but enhancing knowledge density through technological innovation. This will enable AI models to do more with fewer parameters and be deployed on more intelligent devices. The goal is to make intelligent personalized services accessible to all.

How can we change on-device cloud-based AI from being merely usable to being user-friendly and transformative? How should we view the industry trend of synergy across device, edge, and cloud?

Liu Zhiyuan: Moving forward, AI is predicted to evolve along two major tracks. The first is making models more intelligent and more capable. This will turn them into autonomous agents that can learn and explore independently within their designated domains, in which they can eventually become experts. The next step is to enable communication and collaboration among models so that they can proactively harness collective knowledge to deliver better services to human users. The second track is improving model efficiency to lower model development and usage costs. This will allow quality large models to be deployed on diverse intelligent devices, which will in turn make them more intelligent. At the same time, models should be brought closer to end users by being deployed on intelligent devices nearer to them—closer to user data and user needs—to be more effective at serving humanity.

How do you view the synergy between cloud-based large models and on-device intelligence, and how can they work together to unlock greater potential?

Liu Zhiyuan: After 80 years of development, mobile phones, PCs, and automobiles now possess high levels of computing power. The same is true for cloud computing servers. In the intelligent era, a collaborative mechanism will emerge between intelligent models on devices and those on the cloud to form device-edge-cloud synergy.

What are the characteristics of cloud-based models? They possess abundant computing power, which enables them to provide top-tier intelligent services across a wide range of professional fields.

Devices are closest to users, so they can act as personalized assistants for every user as they continually gather user data, understand user needs, and intelligently tap into cloud-based professional resources to serve users better.

Looking ahead, devices and the cloud will complement each other: Cloud-based models will provide top-tier intelligent services, while on-device models will continually identify and interpret users' deep-rooted and always-present needs. In the future, a mutually reinforcing relationship will be built between the cloud and devices.

Interview 5: How China Unicom is powering the Mobile AI era

Li Hongwu, President of China Unicom Research Institute

It has been almost six years since China launched 5G, and China Unicom has done a great job building its 5G networks. Given the new network requirements being raised by AI, what are China Unicom's plans for 5G-A and what actions have you taken?

Li Hongwu: We are ramping up efforts in 5G-A network construction to keep pace with rapidly evolving technologies and meet the growing demand for better and differentiated experiences in the AI era.

So far, we have deployed more than 300,000 5G-A cells in over 300 cities, focusing on the large-scale deployment and application of four key 5G-A capabilities: ultra-high speeds, intelligent connectivity of everything, deterministic reliability, and integrated sensing and communication. Let's have a closer look at them one by one.

First are ultra-high speeds.

The 3.5 GHz band is known for its large 300 MHz bandwidth, while the 2.1 GHz band can help improve uplink performance. By taking advantage of these two bands and applying technologies like multi-frequency coordination and carrier aggregation, we can deliver a downlink peak rate of 5 Gbps and an uplink peak rate of over 1 Gbps. These ultra-high speeds lay a solid foundation for us to deliver new immersive services like AI agents, embodied AI, and HD live streaming anytime and anywhere.

Second is intelligent connectivity of everything.

We have launched RedCap services in more than 300 Chinese cities and RedCap is now widely used in scenarios like IoV, smart grid, industrial Internet, and video monitoring. This shows how we have already made progress in RedCap commercial use across all frequency bands, RATs, scenarios, and industries.

Third is deterministic reliability.

By leveraging our leading capabilities in 5G and industrial Internet, we have achieved what we call device-network collaboration. This progress guarantees ultra-low latency and reliable data transmission in demanding scenarios, which will further promote the integration of 5G-A into core enterprise production activities.

Fourth is integrated sensing and communication.

We have dived deep into integrated network technologies by combining low-altitude communication networks, sensing networks, navigation networks, intelligent computing networks, and meteorological networks (systems of interconnected weather stations that share data). This kind of integration is helping us build low-altitude "space–air–ground" information infrastructure.

We have currently launched low-altitude service showcases in multiple provinces across China. For example, in Beijing, we released a series of "5G Capital" innovations and deployed the world's first large-scale integrated 5G-A intelligent network. This network enables us to deliver innovative services in cutting-edge fields like smart city services, smart sports, and the low-altitude economy. Furthermore, we have achieved continuous 5G-A coverage across Beijing's main urban areas. This world-leading network has an ultra-large capacity and can deliver a downlink peak rate of 11.2 Gbps and an uplink peak rate of 4 Gbps. We have also deployed the industry's first 5G-A 10 Gbps integrated air-ground network and a smart operations system for the Great Wall. Our experience in these projects can be used as a reference for other carriers looking to achieve success from technological innovation to joint ecosystem development.

A mobile AI era is unfolding as 5G-A meets AI. What are China Unicom's expectations and plans for this new era?

Li Hongwu: To achieve in-depth synergy between 5G-A and mobile AI, we need to build a capability system covering the entire industry value chain. This capability system should include basic network capabilities, reasonably laid out computing power, innovative business models, and ecosystem-wide collaboration.

Specifically, we need to first establish a solid foundation by building an integrated intelligent computing network. This means we must speed up integrated digital infrastructure construction and continuously improve our 5G-A network capabilities. In addition, we need to systematically plan and deploy computing resource pools and build more efficient computing networks. We need to focus not only on cloud-edge synergy, but also on enabling ubiquitous computing power. This will be the key to building more integrated and intelligent communications networks.

Second, we need to create new value by employing AI technologies to reshape our core businesses. For consumers, we should shift from simply monetizing traffic to monetizing experiences by providing more diverse value-added services like computing power and storage. For enterprise customers, we need to be more than just a basic connectivity service provider, and start offering high-value integrated services including integrated solutions, data enablement, and integrated sensing and communications.

Third, we should expand our circle of partners. All industry players across the value chain—whether they are in communications AI, computing power, or applications—need to work closely and integrate their services if we want to build a truly prosperous mobile AI era. The communications industry has a key role to play, acting like a hub connecting users, devices, and data. This means we have a responsibility to build alliances where various industry partners can share resources and complement each other's advantages. They can also work together to make key technological breakthroughs, accelerate innovation, and promote application implementation across more scenarios.

Interview 6: From automation to intelligence: exploring new opportunities in unmanned logistics

Guo Shuangqing, Assistant Chief Marketing Officer, SF Technology

Could you share some of the practices SF has employed and achievements it has made during the digital and intelligent transformation of the logistics industry?

Guo Shuangqing: We started out with express delivery, but have steadily expanded our business portfolio. Today, we focus on comprehensive supply chain services, and we have applied AI technologies to every aspect of these services. For example, our intelligent assembly lines can automatically identify and package goods. When cameras installed five meters from the end of the line capture individual goods, appropriately-sized boxes will be moved to the designated area at the end of the line within one second, right before the goods arrive. This cuts a lot of costs, because with manual operations, workers tend to choose slightly larger boxes to ensure goods fit. However, larger boxes mean larger transport vehicles and thus higher costs, and can particularly cause exponential cost increases for air transport. That's one example of AI's value in the loading process. We also employ AI in after-sales services. For instance, AI models have enabled virtual human services during live streaming. This helps us provide higher-quality and more people-centric services to customers. In a nutshell, SF has applied AI to every aspect of our supply chain capabilities, and shared these capabilities with our enterprise customers.

What challenges do you face and how are you coping with them during the digital and intelligent transformation of the logistics industry?

Guo Shuangqing: Logistically, supply chain operations follow the sequence of collection, transfer, transport, and delivery. From the resource perspective, a supply chain consists of people, vehicles, goods, and warehouses. In recent years, there has been remarkable progress in terms of supply chain automation, which aims to use hardware and software integration to enable autonomous equipment operations, such as automatic lifts and stackers. Now, we are working to tackle the new challenge of intelligent transformation. For now, intelligent equipment, like AGVs, automatic forklifts, and automatic “work bin” robots, is capable of smoothly performing one task in simple scenarios. But in complex environments, a single type of equipment is unable to handle all tasks, and a combination of different types of equipment is needed. Currently, we face two major challenges: First, existing automatic equipment still relies on Wi-Fi, rather than 5G/5G-Advanced networks; second, public cloud platforms are still being constructed. This presents both major opportunities and challenges ahead.

SF has studied logistics drones for 13 years, but we are still exploring how to transform R&D results into commercial applications. In mobile communications, errors are allowed to a certain extent; signal coverage can be overlapped; and breakpoints can exist between base stations. But this is not the case for logistics drone communications which requires real-time connectivity, while information transmission breakpoints or latency will cause problems. We hope that telecom carriers, the government, and related large enterprises can work together to strengthen infrastructure, so as to reduce long-term operational costs and help China develop by leaps and bounds in the low-altitude logistics industry.

The impending 5G-Advanced era is creating opportunities for supply chain logistics, but is also bringing with it the challenge of infrastructure upgrade.