Tech Frontiers: The Road to Innovation for Chinese Companies in the Age of AI

US-China Technology Frontier Series – This Q&A is the second in a series of articles developed by the AI+Web3 Research Center of CKGSB alongside industry experts on the US-China technology frontiers. The series aims to introduce the most cutting-edge theories and practices at home and abroad on the theme of digital technology and internationalization, open up the mindset of students, discuss new business opportunities brought about by the development of science and technology, and identify new opportunities for the technologization and internationalization of Chinese enterprises. You can find the first article here.

Interviewer: Professor Sun Baohong is currently Dean’s Distinguished Chair Professor of Marketing and Head of the Web3+AI Research Center at CKGSB

Interviewee: Larry Zhou is a Fellow and Chief Network Architect at AT&T. In his forward-looking research, Larry focuses on on a number of topics, including Artificial Intelligence, the Internet of Things, Blockchain and Web3. Larry has demonstrated exceptional technical and architectural vision in the telecom industry as a true innovator and industry disruptor.

The development and application of cutting-edge trends in technology—in particular AI, 5G, Web3 and the metaverse—are key to continued global economic development. In this interview, Larry Zhou, Fellow and Chief Network Architect at AT&T, discusses the centrality of 5G technology in driving the development of AI and the metaverse, how open network technology (O-RAN) is enablng SMEs to compete in the 5G market through the breaking down of the current market monopolies, and how the rise of Web3 is paving the way for the assetization of digital content and combining it with AI to open a new chapter in the creation of virtual worlds.

The interview also provides insights into the importance of computational power to AI and other future technologies, and how the distribution of computing power, from cloud to edge computing, affects user experience and privacy protection.

Q. What is the significance of Chinese firms embracing open networks and what is the impact of these networks on 5G and AI technology development?

The next five to 10 years will continue to see the development of 5G technology but, and perhaps more importantly, it will also see the release of wireless communication technologies such as Wi-Fi 7. Without advanced wireless communications technology the rollout of AI lacks wings, because future mobile terminals and associated equipment absolutely require 5G support.

But I do not want to discuss 5G in the traditional sense, rather two new concepts: open networking and open source. One example of open networking [software or infrastructure designed to create more flexibility, interoperability and automation for hardware from different vendors] is the Open Radio Access Network (O-RAN), which can basically be understood as a newer type of telecommunications base station. China currently leads the world in the number of 5G base stations as well as the technology itself, and open networking is of great significance in the development of 5G.

The idea of O-RAN was jointly put forward by AT&T and China Mobile to help break the monopoly over base stations held by companies such as Huawei and ZTE in the China market, as well as Ericsson and Nokia in the international market. This dominance made it very difficult for smaller tech companies to enter the market, but O-RAN has shifted this status quo, enabling small- and medium-sized enterprises (SMEs) to participate in the development and application of base station technology. In fact, it is likely that these SMEs will be the drivers of O-RAN development, given that the larger incumbents will embrace the technology less aggressively as it threatens their existing market share.

While the concept of O-RAN was introduced initially in 2017, it wasn’t until 2024 that AT&T officially adopted the solutions on a large scale, signing a five-year $14 billion contract with Ericsson. The open architecture allows the cost of base station equipment to be significantly reduced, while improving the versatility and interconnectivity of the equipment. In turn, this helps with more flexible deployment, better promoting the widespread coverage of 5G.

O-RAN’s openness enables more companies to participate in the development of base station equipment, reducing deployment costs and helping accelerate 5G coverage—all things that boost AI applications and the development of the metaverse, as each requires the low latency, high privacy and high security provided by 5G. AT&T’s plan to shift 70% of its wireless network traffic to O-RAN by 2026 is an effort to drive the growth of this open network.

To futher accelerate the rollout of O-RAN or other network technologies, I put forward the idea of a ‘White Box’ for network virtualization [the process of transforming network functions into software, and disconnecting them from the hardware that they traditionally run on]. The core idea of the concept is to virtualize network equipment and run it on an open hardware platform, rather than relying on proprietary equipment provided by large companies such as Cisco or Huawei. This open architecture has not only accelerated technology upgrades, but has also dramatically reduced costs, making high-speed networks ubiquitous.

It is clear that the development of open networks is important to the advancement of AI, the metaverse and the future of cyber infrastructure. The network of the future will be a convergence of wireless and wired, where switching between different networks will be seamless for users, whether it’s from 5G to Wi-Fi 7 or to a wired network in the home. The shift towards network convergence will drive the further development of communication technologies globally.

Q: What is the relationship between Web3, the metaverse and the development of AI?

My generation has witnessed and participated in the development of the Internet, from its earliest days to today’s Web3. In the early days of the Internet, the Yahoo era, for example, users were only really able to receive information passively, i.e. reading content but not creating their own. With the emergence of Web2, platforms like Facebook allowed users to upload their own content and share it with others, providing an interactive function, like graduating from simply reading, to reading and writing. And now we’re entering the era of Web3.

Web3 not only allows users to read and write content, but also gives users copyright and control over the content they upload, making it a digital asset that the user owns. Web3 also means we can generate a ‘Spatial Web’ of interaction between the digital and physical strands of our lives, creating a three-dimensional interactive experience that will greatly enhance the user experience compared to the flat communication of the past.

With the rapid development of AI, the future of Web3 will not only be about human interaction with the web, but also interactions between AI and web content. This ‘Semantic Web’ will enable AI to understand and process content from Web3 sources, providing benefits to us and to the AI itself.

There are many Web3 applications that have gained notoriety around the world, such as Bitcoin, Blockchain, NFTs, Decentralized Finance (DeFI) and so on. And these technologies and applications form a solid foundation for the further development of Web3, and also play an important role in the construction of the metaverse. Additionally, since the emergence of GPT-4 and other recent AI tech, the application scenarios for the metaverse have massively expanded.

As an example, in the future the metaverse will undoubtedly be home to a huge number of virtual characters, also known as non-playable characters (NPCs), and thanks to AI support, these NPCs will not only resemble humans in the real world, but they will also have their own intelligence and emotions. NPCs could potentially form their own social structures, and humans may see them as friends or confidants when interacting with them. It’s even possible that our closest friend or companion could be a relationship formed in the virtual world.

Leaning even further into what feels like sci-fi, you may even be able to clone yourself in the metaverse, creating a virtual avatar that is identical to you in the way it looks, behaves and thinks. Although this technology sounds fantastical now, the progress of AI and metaverse technology means that such a scenario is gradually becoming possible.

The relationship between Web3, the metaverse and AI is both complementary and mutually reinforcing. For example, AI can generate a great deal of content, such as 3D models, in the metaverse, which greatly accelerates the creation of the metaverse itself. In turn, NFTs and cryptocurrencies in Web3 provide the necessary infrastructure for the establishment of an economic system in the metaverse, allowing virtual assets to circulate.

Web3, the metaverse and AI are technologies and ideas that fuel each other, and together they drive the construction and development of the future digital world.

Q: What are the current and potential applications for AI in robotics?

Robots are evolving from mere mechanical actuators to intelligent assistants. For example, RT-2 (Robotic Transformer 2), which combines the Large Language Model (LLM) and the Diffusion Model, has demonstrated visual-linguistic-action (VLA) synergy. Industrial robots can be used in warehouses, docks, and production lines, while domestic robots can help the elderly improve their independent living skills, encompassing health monitoring, assisted living and emotional companionship.

Q: What impact will brain-computer interface technologies such as Neuralink have?

Neuralink enables AI to read and understand human brain signals directly in order to perform tasks. For example, it could allow people who have lost their eyesight to see the world again, or enhance human cognitive abilities through a “second brain.” Future applications of this technology will dramatically change the way humans interact with AI.

Q: To what degree is computing power important and how do you think the industry will evolve in the future?

Computing power plays a crucial role in current and future technological development. None of AI, the metaverse or Web3 could evolve without strong computational support. We need a lot of computing power when we train AI models and the same when running large language models or multimodal models for inference. In addition, the rendering of 3D models in the metaverse, particularly with the need for low-latency rendering, is also highly dependent on computing power. Even cryptocurrency transactions and verification in Web3 require computational support.

The invention of the transistor by AT&T in 1948 was an important milestone in human history, replacing the bulky vacuum tubes that were used before then. The creation of transistors is what made chips possible, which in turn gave rise to devices such as cell phones and modern computers.

The core of computing power is the semiconductor chip, and, unfortunately, China’s chip manufacturing industry is currently facing significant barriers to the creation of cutting-edge chips. The performance of a chip depends on the number and size of its internal transistors, and as the size of transistors continues to shrink (e.g. 7nm down to 3nm etc.) more transistors can be integrated into the same chip, thus improving its performance.

Chips are critical to the development of AI, but we don’t necessarily need state-of-the-art chip technology for AI—larger versions can be used to achieve the required functionality. So, even if China can’t break through to the 3nm level just yet, it is still possible to use 7nm chips to fulfill AI’s computational needs, as well as make larger GPUs or TPUs for AI and the metaverse. While cell phone chips require smaller transistors due to space constraints, this limitation is not significant for AI and the metaverse.

Another solution is the miniaturization and distributed process of AI models. We can deconstruct a larger model into multiple smalle models running on smaller chips. With this distributed computing, multiple smalle models can work together to achieve similar results to the larger model. By doing this, the computational requirements for training and inference can be drastically reduced, avoiding a dependence on high-end chips.

In terms of distributing computational power, the main focus is currently in cloud computing. However, edge computing [which brings computation and data storage closer to the user, rather than being entirely reliant on a centralized data centre] is becoming increasingly important. In AI and metaverse applications in particular, the closer the edge computing is to the user, the better the user’s experience and privacy protection will be—storing and processing all data in the cloud can increase risk of privacy breaches.

Cloud computing and edge computing will continue to develop together, but I think the latter will do so at a faster rate. In a course at Princeton University, they put forward the view that unity breeds division, and this also applies to the centralization and decentralization of computing power. In the future, with changing application requirements, the distribution of computing power will gradually shift from the cloud to the client, and edge computing will play an increasingly important role.