On November 13, the 2024 Global 6G Development Conference opened in Shanghai. The combination of 6G and artificial intelligence has become a hot topic discussed at this conference. Wu Hequan, an academician of the Chinese Academy of Engineering, pointed out in the keynote speech that we need to attach great importance to the impact of artificial intelligence on 6G systems, but it is not appropriate to be too optimistic at the moment. From his personal point of view, there are still challenges in the application of AI in wireless access networks. In the next few years It is difficult to predict how much revenue it will bring each year.
Photo courtesy of organizer Wu Hequan, academician of the Chinese Academy of Engineering
6G is the sixth generation of mobile communication technology, which supports higher transmission rates and extremely low latency. Compared with 5G, 6G proposes new scenarios, such as the combination of communication and perception, the integration of communication and artificial intelligence, and ubiquitous things. Connection refers to information connection and interaction between anyone and anything at any time, anywhere, and 6G will be more adaptable to the arrival of the artificial intelligence era.
Wu Hequan believes that the discussion about artificial intelligence is very hot now. It is true that artificial intelligence needs to be combined with 6G, and 6G also needs artificial intelligence even more. In terms of application, the current basic large model itself can be used in 6G, especially intelligent customer service and some supply chain management. Some scenario models can also be developed for network planning and optimization.
However, at this stage, AI is facing great challenges in the application of access networks, because wireless access networks require high real-time performance. The calculation and processing of large AI models take up time and there is a certain delay. Whether it can respond in real time is a question. A big challenge. "Especially on urban roads, if real-time convergence is not possible, the effect may not be very good; secondly, it will definitely increase complexity. Currently, the interference of a single cell is OK. In the future, we hope to use AI to offset the interference between multiple cells. However, when calculating multiple cells, the computing power requirements are very high. "Wu Hequan said that although it is conceivable that AI can play a role in counteracting multi-cell interference, whether it can actually play a role is also a big problem. challenges.
One thing that needs to be considered is that artificial intelligence applications require investment costs. Whether the returns obtained are enough to balance the costs needs to be measured. In addition, the current large models themselves have insufficient interpretability, and some may not necessarily form credible solutions and require human testing.
However, Wu Hequan also pointed out that the current new solution is to compress large models onto mobile phones through model quantification, model editing, etc. When this reasoning ability, including the corresponding computing power, sinks to the terminal, there is no need to go online. Cloud, but applied offline with low cost, low latency, high privacy and personalization. He believes that this will help enhance the ability of users to generate content and drive the development of a new generation of information services industry.
He also mentioned that the emergence of the 4G era has driven the development of short videos, but in recent years, "basically there have not been many changes in the baseband processing of mobile phones. What has changed are changes in capabilities such as photography." He said optimistically: "In the future, I am optimistic that AI will sink to the terminal and drive the development of user-generated capabilities."
Wu Hequan emphasized that the development of 6G should pay more attention to the actual needs of the public, rather than just pursuing high indicators in niche scenarios. "The problem now is that our standards for 6G tend to only pay attention to those specific requirements and put forward a bunch of high indicators for specific requirements. I suggest that when 3GPP studies standards, it should study some reasonable requirements specifically for the public who need it. , the main 6G system design is first aimed at serving the requirements that are not so high but satisfy the public experience. "
In his view, although technologies such as holography, sensory interconnection, and 3D immersive experience can reflect the broadband level of 6G, they are "not a rigid need of the public, and they are not an application for handheld terminals." He further explained, "Nowadays, the screen of mobile phone terminals is not very big. There is not much difference between 2K, 4K and 8K, and it is not suitable for viewing in sports conditions." Wu Hequan called on the industry to promote the development of 6G more pragmatically, "Future 6G terminals should is diverse, general terminal We only need to support the public’s urgent needs. For some applications with specific requirements, we can obtain computing power through short-distance communication via mobile phones to fixed networks or edge computing.” He emphasized that the industry should prioritize meeting the actual needs of the public to ensure that 6G technology is truly real. Serving the public.